An Agent-Based Sustainability Perspective on Payment for Ecosystem Services: Analytical Framework and Empirical Application

Abstract

Payment for Ecosystem Services (PES), a market-based policy instrument for the conservation and environmental management that aims to coordinate the interests of upstream and downstream ecosystem service (ES) stakeholders, has been adopted worldwide. However, the success of PES depends on the desirability of programs targeting rural communities and smallholders. In this article, an agent-based sustainability perspective on PES was proposed and applied to examine a PES case study of the Converting-Orchard-to-Forest (COF) project in Dongjiang Headwater Watershed (DHW). We used household interview-based information and associated secondary data to quantitatively assess the environmental consequences and livelihood impacts of the COF project. The findings show that: (1) the COF participants at the upstream suffered from substantial income loss due to decreased orchard area; (2) the participants’ chemical fertilizer and compound fertilizer consumption was larger than their nonparticipating counterparts; and (3) the COF participants and nonparticipants increased the material assets and reduced their fuelwood use and increased the liquefied petroleum gas. Our findings suggest that, because of the significant income loss experienced by the upstream participants, the COF program is unsustainable with the participants very likely to cultivate the orchard again once the COF project ends. The research provides insightful information regarding PES implementation and sustainability of similar PES schemes.

1. Introduction

Payment for Ecosystem Services (PES) as a promising policy instrument tries to translate the positive externality of non-marketable natural systems into an economic incentive of ecosystem services (ES) providers, such that they will strengthen conservation efforts and provide environmental public goods [1,2]. In watershed-based PES projects, downstream water users typically provide financial compensation directly to upstream communities for water resources management and conservation efforts to adopt land-use decisions that are assumed to prevent deforestation and enhance hydrologic services [3,4,5,6,7,8]. There are three main types of PES schemes—user-financed PES, government-financed PES, and compliance PES—depending on the participating community and state involvement [9,10,11]. PES schemes emphasize the importance of governmental support, as evidenced in Costa Rica, Mexico, Nicaragua, and Colombia [12]. Current PES schemes aim to understand how to cultivate the implementation to improve efficiency and effectiveness and address tradeoffs to secure human well-being [13,14,15,16,17]. An application of the Payments for Watershed Services-Watershed Sustainability (PWS-WS) framework illustrates the potential consequences of relying on indicators of the socioeconomic and governance dimensions using semi-structured interviews and focus group discussions with farmers when evaluating PWS program performance [18,19].

A large number of watershed PES initiatives have been successfully implemented globally, such as the world’s largest PES Program, the Grain to Green Program (Sloping Land Conservation Program), and the Natural Forest Conservation Programs in China, and a central PES implementing authority in Costa Rica and South America (Pagos por Servicios Ambientales, PSA), Ecuador of Pimampiro with a decentralized authority, Bolivia and Mexico’s PES schemes [20,21,22,23]. PES schemes are believed to not only benefit ES providers [24,25] but also improve the beneficiaries’ life quality by delivering both environmental and social advantages, which in turn advance the PES implementations [26]. Specifically, the PES designs need to effectively meet the biophysical objectives, including serving the beneficiaries, allocating resources, and improving transparency during the processes [6], thus affecting the programs’ success/sustainability in the long run [21]. However, the results from prior literature regarding such effectiveness are mixed [27,28]. One possible reason could be data limitation. The other reason, which can be more significant, is that extant studies tend to focus on the environmental effects alone while ignoring the social impacts although the latter could disincentive the ES providers, thus leading to PES’s inevitable failure [9]. A few exceptions studied both the environmental and social dimensions of PES. Yet, their findings are inconsistent regarding if there is a synergy between the two and, more importantly, under what conditions [29].

Participants’ livelihood changes depend on PES characteristics and context and will inevitably bring socioeconomic and environmental effects; therefore, the evaluation of livelihoods changes is crucial towards PES sustainability and cost-effectiveness [30]. Lacking adequate financial support associated with PES institutions and regulations is still the main obstacle for PES sustainability [26]. A sufficient incentive payment to landowners should be high enough to cover opportunity costs to change their land-use behavior [9]. A deeper understanding of these schemes’ costs would be valuable in assessing cost-benefit relationships and designing new or improving existing PES programs. Although many scholars identified the cost-benefits of the PES program targeted to enhance economic efficiency, the information on participants’ livelihoods changes remains unclear [12,31]. Using the Paddy Land to Dry Land (PLDL) program as a case study to incentivize upstream farmers through direct payments to adopt land-use practice using household survey data, Zheng et al. (2013) quantified associated costs-benefits tradeoff to both ES providers and beneficiaries separately indirect effects on socioeconomic and environmental dimension and the following livelihood activities changes [32]. Several institutional structures of PES including income sources and benefit distribution system, funds utilization efficiency, and policy are believed to determine the success of PES schemes, such as high payments, a high degree of voluntary participation, low transaction costs [18].

However, PES cannot be regarded as a priori as the most cost-effective policy to achieve desired ES targets [11,33]. The environmental and socioeconomic impacts and institutional, political context of PES schemes can be understood by assessing observation data from water quality and hydrological monitoring networks [29,34]. As for most PES schemes, we have no clear knowledge about the effectiveness [9]. Previous PES research indicated very mixed conclusions for PES effectiveness. PES projects lack baseline data and randomized design, proposing difficulties in achieving a comprehensive effectiveness assessment. Several studies depend on secondary data, likely creating additional uncertainties influencing the PES assessment results due to the uncertainty of hydrologic measurement data [2,15]. Some research depends on case studies, causing selection errors. The framework provides a useful approach for guiding interdisciplinary efforts to enhance understanding of the complex drivers and feedback that determine the potential to achieve long-term goals related to hydrologic services while positively influencing the well-being of upstream and downstream communities [24,35]. The framework of telecoupling can help promote sustainable, systematic, multidisciplinary studies on different types of distant interactions and their interrelationships [21]. PES effectiveness assessment is to apply the institutional framework, and there is no comprehensive framework existing to date. Therefore, in-depth knowledge of the balance and tradeoff between benefits and costs resulting from participating in PES projects is crucial for understanding whether ecological restoration projects can improve ecosystem services. In this research, household surveys and statistical analysis were applied to probe the household livelihood changes of participating households and nonparticipant households from 2010 to 2015 and address the ecological effects of a PES project in China and the environmental consequences for the upstream and downstream stakeholders. Therefore, the study’s objectives are to develop an integrated framework and apply the Converting-Orchard-to-Forest (COF) as case studies to enhance PES’s effectiveness, both environmental improvement and livelihood changes to achieve PES sustainability.

2. An Agent-based Analytical Framework for Assessing PES Sustainability

An appropriate institutional framework has been found critical for management control and payment distribution arrangements and the success of community-based forest enterprises [36]. Agent-based modeling (ABM) simulates the effects of various social and environmental factors on the emergence and evolution of social norms. It is widely used for research in land changes and coupled human and natural systems [37]. ABM is a bottom-up approach that predicts emergent higher-level outcomes by simulating individuals’ decision-making and their interactions with each other and with their environments. Chen et al. (2012) examined the effects of social norms on enrollment in China’s Grain-to-Green Program. Zheng et al. (2013) applied a PES provider-beneficiary’s framework. They found that the PLDL program has led to changes in livelihood portfolios, household production and consumption activities, and the export of nutrient waterways. Applying the telecoupling framework to PES programs can help identify research gaps and standardize analytical approaches with flexibility based on specific contexts [21]. Payment for Watershed Services (PWS) as a policy tool for enhancing water quality and supply has gained momentum in recent years. Still, their ability to lead to sustainable watershed outcomes is uncertain. Asbjornsen et al. (2015) developed a new PWS-WS framework to guide indicator selection to improve knowledge about the complex drivers, interactions, and feedback between PWS and Coupled Human and Natural Systems (CHANS) [18]. Anderies (2015) presented a general mathematical modeling framework that can provide a foundation for the study of sustainability in social-ecological systems (SESs) [36]. The telecoupling framework emphasizes interactions among sending, receiving, and spillover systems via energy, matter, and information flows [21]. The telecoupling framework could be used to assess the synergies for PES schemes among the three integrated systems. The difference between the coupled infrastructure systems (CISs) framework and other telecoupling frameworks is that the former emphasizes the sustaining ecosystem service provision, balancing the tradeoff between costs and benefits, and proposing regional cooperation to approach three wins’ outcomes among upstream, downstream, and whole ecosystems. Payments for Watershed Services (PWS) Watershed Sustainability framework integrates biophysical and socioeconomic indicators to assess progress toward watershed sustainability goal (PWS-WS) [18]. This framework could guide indicator selection to improve information about the interactive drivers and feedback between payments for watershed services and the coupled human and natural systems. The difference between CIS and PWS-WS is that the latter framework focus on the changes in ecological policies upon the biophysical factors. The framework is designed to pose direct impacts on forest cover and hydrologic services. Considering equity, the PES framework’s core components are essential factors to prevent environmental degradation in management practice [24]. The framework emphasizes the dimensions of equity.

We presented an agent-based coupled infrastructure system (CIS) framework for assessing PES sustainability (Figure 1). The framework highlights three points. Due to non-marketable goods/services produced by a focal CIS and maintaining the optimal provision, ES as positive externalities has given rise to dedicated market-like arrangements. PES schemes should have environmental externalities of human actions internalized. The externalities often lead to a market failure, and stakeholders will incline to under-provide ES [2]. Incentive-based policies take externalities through the changes of economic incentive and management practices. The CIS’s upstream coupling affects decision-making and transfers goods and services via PES schemes to downstream coupling stakeholders. The framework was extended based on the “optimal control” conceptualization of socio-environmental systems (i.e., CIS) by Anderies (2015), through highlighting three points: (1) PES schemes as a marketization means to internalize the externalities due to the non-marketable good/services generated by a focal CIS; (2) decentralized decision-making of the ES suppliers in the focal CIS, concerning production allocation of eco-environmental infrastructures and socio-economic infrastructures as well as to the consumption of marketable and non-marketable goods/services; and (3) the focal CIS’s downstream coupling affecting the ES beneficiaries as well as its upstream coupling that affects the ES providers’ decision-making. This recognition of spillovers’ importance suggests that we need to move from SESs to coupled infrastructure systems (CISs).

3. Empirical Case and Research Methods

3.1. The Research Context

The COF project was implemented in the Dongjiang Headwater Watershed (DHW, including Xunwu, Anyuan and Dingnan County) in southeast Jiangxi Province of China, which is a significant drinking water source of Pearl River Delta (including Guangzhou, Shenzhen, and the Hong Kong Special Administrative Region, SAR) (Figure 2A). The watershed area is 3584.64 km² with one million permanent residents, one-third of them fall below China’s poverty line. The percentage of forest land in all land-use types is approximately 74.18% [38]. Since the 1980s, many farmers have raised their income and improved human well-being through orchard plantations. Since DHW is located in a subtropical moisture monsoon climate zone with four pronounced seasons, its biophysical conditions are especially suitable for cultivating citruses orange and navel orange in the hilly area.

The terrain with simple modification would be turned into a high-quality orchard garden. Since the 1980s, the orchard industry has gained apparent achievements. The orchard development in DHW has been regarded as the key industries. The orchard area in Gannan (southern part of Jiangxi Province, DHW is a part of Gannan) has reached 2.60 million mu (1 mu = 1/15 ha), ranking the first position in the world. The orchard quantity in 2014 is 3 million tons, and the industry value was 8 billion yuan (1$ = 7.2 yuan) in 2014. Many of the farmers, through orchard production, have raised their income and improved local human well-being. DHW is known as the “hometown of Chinese orchard citrus,” with its annual orchard production accounting for 44% of the total national quantity [38,39,40].

The fertilizer application has contributed to soil acidification and agricultural non-point source pollution. The COF project pays people to cut down their orange trees voluntarily to restore critical ecosystem services and enhance drinking water quality for downstream communities. One of the most significant challenges is how to balance the interests and livelihoods of both upstream and downstream stakeholders. DHW is a vital drinking water source area and provides 2.90 billion tons for the downstream (Shenzhen, Guangzhou, and Hong Kong SAR). The water quality and quantity have critical implications for the drinking water safety of 40 million residents in Pearl River Delta (PRD) and Hong Kong SAR. It provides about 0.87 billion tons of water for Shenzhen (accounting for 66% of the total water demand quantity), 1.10 billion tons for Hong Kong SAR (70% of the water demands), and 0.40 billion tons for waterfront areas of Dongjiang. Unfortunately, before implementing COF in 2010, the rapid expansion of orchard plantation had severely degraded the water quality, which raised growing concerns from the downstream Guangdong Province and Hong Kong SAR [41]. In 2012, the average water quality below III standard accounted for 38% of the water in Xunwu River, Dingnan River, Mati River, and Xiali River. As Jiuquwan Reservoir was an example in 2012, the total nitrogen (TN) mean value exceeded the lake water quality standard II by 1.53 times. The total phosphorus (TP) mean value was by 0.81 times.

Under the help of modern agricultural tools, much of the steep slope land above 15° in 2010 was converted into an orange plantation with an area of 48,000 ha, accounting for 33% of the total orange plantation area [38]. According to the Technical Specification for Navel Orange Plantation, the orchard slope should not exceed 15°. Otherwise, it will cause severe soil erosion and even mud-rock flow. The soil erosion area in 2012 is 97,752.70 ha, and about 92.43% of orchard planting area have appeared soil erosion. The water-soil loss with medium intensity accounts for 42.85% of the total area, and the soil erosion modulus is 4280.00 t/km²∙a [42]. The annual soil loss amount is 4.18 million tons, of which the surface soil has been eroded of 0.2–1.0 cm annually with plenty of organic matter and N, P nutrient. The farmers invest 1080.00 kg of compound fertilizer, 297.00 kg of carbamide, and 11,812.50 kg of organic fertilizer per ha [43]. The chemical fertilizer utilization per year is 369.47 kg/hm² and is obviously above the standard of 280.00 kg/hm² for ecological towns in Jiangxi Province [44]. The continuous soil organic matter (SOM) content decreased to 7.43 g/kg for the first year orchard plantation, which is drastically below the normal average SOM content of 13–15.00 g/kg [38]. Moreover, large-scale orchard development has caused the landscape singleness, ecosystem function deterioration, and decreased the resistance ability to disease and insects.

The Horizontal Ecological Compensation Agreement between Upstream and Downstream of Dongjiang was signed by Jiangxi Province and Guangdong Province that focus on pollution treatment, ecological restoration, water source conservation, and water and soil erosion management. Simultaneously, the Guangdong provincial government has paid a maximum of a total of 0.2 billion yuan to compensate the Jiangxi local government in terms of water quality and quantity and other environmental monitoring projects. In this context, a Trans-boundary Payment for Ecosystem Services project (the COF project) was implemented in 2010, involving mainly the Guangdong provincial government and voluntarily participated in rural households in the three DHW counties. Specifically, a mountainous region above 25° along the Xunwu River has been defined as forbidden development zones. Participating rural households converted their orchard land to broadleaf forest plantation and mangosteen for the purpose of environmental restoration (Figure 2E). The COF participating and nonparticipating rural households were surveyed. The orchard area is 50,427 ha in 2010 and reduced to 45,527 ha in 2015 due to the COF implementation. Approximately 19 million trees in the high production period have been cut off, which has reduced the local government’s financial income and posed adverse effects for farmers’ living level.

3.2. Methods

3.2.1. Method for Estimating the Livelihood Impacts of the COF Project

Propensity score matching (PSM) constructs a statistical comparison group that is based on a model of the probability of participating in the treatment, using observed characteristics. This matching can help strengthen causal arguments in quasi-experimental and observational studies by reducing selection bias [45]. PSM is to evaluate the policy’s effects consisting of a treatment group and control group [45]. The participation group is the treatment group and the other is the control group, reaching the balance between the two groups through the propensity score and avoiding selection bias. The main steps for PSM include (1) estimates a model of program participation; (2) define the region of common support and balancing tests; (3) matching participants to nonparticipants; (4) calculate the average treatment impact. Propensity score methods are listed as follows. After matching on propensity scores, we could compare the outcomes of treated and control observations.

$$ATET=E(\Delta |p(x),D=1)=E(y_1|p(x),D=1)-E(y_0|p(x),D=0)$$

(1)

where ATET is the average treatment effect on the treated and is the difference between the outcomes of treated and the outcomes of the treated observations if they had not been treated. D = 1 represent the treated observations and D = 0 for control observations. P(x) is the propensity of observation to be assigned to the treated group. X variables may affect the likelihood of being assigned to the treated group.

Each treated observation i is matched to a j control observations and their outcomes y₀ are weighed by w.

$$ATET=\frac{1}{n_1}{\displaystyle \sum _{i\in \{D=1\}}\left[y_{1,i}-{\displaystyle \sum _{j}w(i,j)y_{0,j}}\right]}$$

(2)

where w is the weight that can be assigned into treated groups.

To assess livelihoods changes resulting from the COF program, the difference-in-differences (DID) approach was used to identify the different responses resulting from COF program implementation. Regarding a public policy is a natural experiment, comparing the treatment group, that has been affected by the impacts of policy, and control group (the control group is unaffected by experiment factors); we can derive the effect resulting from policy implementation [46,47,48]. The sample affected by the new policy is the treatment group, and the corresponding is the control group. It has several advantages. For example, the model is easy to use, and the regression estimation method is maturity. Comparing the static comparison method, the DID method does not directly compare the samples’ mean change during the pre-post policy implementation and it can avoid the pretreatment differences. However, DID methods have some drawbacks, such as the endogenous, the affected control group, and samples’ heterogeneity.

Formally, the standard DID estimate of impact can be denoted by the following:

$$Z_{DID}=[E(Y_t|D=1)-E(Y_{t^{\prime }}|D=1)]-[E(Y_t|D=0)-E(Y_{t^{\prime }}|D=0)]$$

(3)

where $Y$ is the outcome of interest, D represents whether the household is a participant (1) or not (0), $t$ denotes the period when the program is in operation, $t^{\prime }$ indicates the period before the program begins, and $E$ is the expectation operator. DID estimator is the difference in Y for participants across the two time periods minus the difference in Y among nonparticipants over the two time periods [32]. We analyze these changes with two variants of difference-in-difference (DID) techniques, which address changes of participants in the COF program compared to those who do not participate in the program in 2015 (after the project was implemented) versus 2010 (before the project was implemented). The estimates of consistently estimating propensity scores require a dataset to determine program eligibility with independent covariates [32,48] adequately. The DID matching estimator can be expressed as follows:

$$Z_{DIDM}=[E(Y_t|P(X),D=1)-E(Y_{t^{\prime }}|P(X),D=1)]-[E(Y_t|P(X),D=0)-E(Y_{t^{\prime }}|P(X),D=0)]$$

(4)

where P(X) is the probability that a household is selected into the program as a function of household characteristics X. Our results included both ZDID and ZDIDM. The t-test was applied to demonstrate the difference between participant households and nonparticipant households.

In the study, we used the MatchIT package in R to conduct the PSM analysis because R is open-source software and is widely used by data scientists across many different fields [49]. The validity of PSM depends on conditional independence and sizable common support or overlap in propensity scores across the participant and nonparticipant samples. The matching method is nearest. The essence of PSM-DID is to classify the groups using PSM methods and then to calculate the policy effects based on DID methods.

3.2.2. Survey of COF Participating and Nonparticipating Rural Households

To investigate the livelihoods changes resulting from the COF implementation, we conducted the questionnaire in Xunwu County, Anyuan County, and Dingnan County in August 2016, April 2017, May 2017. The survey included questionnaires for rural households and communities, and some semi-structured individual interviews face to face, as follows. A total of 39 villages in 26 towns were chosen randomly within the survey villages. We surveyed about ten randomly selected households from each village. At last, a total of 236 households participating and 103 nonparticipating households were selected randomly. The changes in rural households’ livelihood, earnings, and productive activities, and the diversity of income sources in 2010 before implementing the program and their current status in 2015 were asked. The questionnaire focused on the household level: (i) production and demographic characteristics; (ii) livelihood assets; (iii) changes of income and consumption activities.

Table S1 is a summary of individual characteristics and household demographics of respondents in DHW. It shows the survey respondents were mainly male. The age interval percentages between 30–50 and 50–70 years old are 44% and 51%, respectively. This makes sense since young people in rural areas are likely to migrate to cities to find high-paid jobs, which leaves their hometowns full of old-aged farmers. Most farmers’ average education level among the participants and nonparticipants in DHW is below nine years and do not have college educations. The total income is 85,282 yuan and 93,074 yuan in 2010, which their primary income was from the orchard plantation. They have little arable land to feed themselves. What is important to note is that some of the interviewees in our surveys previously worked as village cadres. We believe that the inclusion of village cadres in our surveys enhances our results’ accuracy since they play essential roles in the rural economy and politics.

4. Results

4.1. Environmental Effects of the COF Project

Due to COF implementation, participant households’ orchard area decreased sharply from 0.694 ha to 0.085 ha in the two periods (Figure 3A). Both the participant and nonparticipants have reduced their orchard area at different levels. This indicated that participants had reduced their income because their orchard trees have been cut down following the government’s policy. While nonparticipant households’ orchard area has decreased slightly from 0.695 ha to 0.442 ha, their income will not dramatically change (Figure 3A). As for the chemical fertilizer and compound fertilizer amounts between participants and nonparticipants used, participants’ chemical fertilizer in 2010 and 2015 decreased from 2053.320 to 1038.520 kg/ha while nonparticipant’s chemical fertilizer increased from 682.840 to 1121.010 kg/ha. The participants’ amount of chemical fertilizer and compound fertilizer amounts in 2010 was larger than that of nonparticipants while the amount was smaller than the nonparticipants in 2015. Meanwhile, the participants’ compound fertilizer decreased from 2098.630 kg/ha to 678.030 kg/ha, and nonparticipants’ compound fertilizer increased slightly from 1145.390 to 1170.870 kg/ha during the two periods. In 2010, the pesticide usage amounts of participants and nonparticipants were 32.230 kg/ha and 32.292 kg/ha, respectively. In 2015, the pesticide amount of participants and nonparticipants were 3.912kg/ha and 20.513kg/ha, respectively (Figure 3B). This indicated that the participants reduced their pesticide usage amounts compared with nonparticipants and achieved the planned environmental goals, i.e., enhancing the hydraulic services and reducing fertilizer amounts.

Moreover, the participant household decreased the chemical fertilizer and compound fertilizer amount. The COF project has demonstrated that it played a crucial role in alleviating water pollution in DHW. Figure 4A showed the changes in rural households’ average orchard yields in 2010 and 2015. Participants’ orchard average yields in each household in 2010 were 29,950.850 kg, and it sharply reduced to 3702.290 kg in 2015. However, the nonparticipants’ orchard yields were 21,976.760 kg in 2010 and 13,490.630 kg in 2015. The main rural non-point source pollution type is COD, NH₃-H, TP, and TN. Figure 4B indicated orchard discharges of Chemical Oxygen Demand (COD), NH₃-N, Total Nitrogen (TN), and Total Phosphorus (TP) in 2010 and 2015. The COF program has reduced COD, NH₃-H, TN, and TP content by 32.650, 313.010, 138.230, 25.580 tons per y, respectively, and improved water quality since the program was initiated in 2010. The orchard area in DHW in 2010 and 2015 is 50,427 ha and 45,527 ha. The pollutant discharge of COD between them is 336.000 tons and 303.351 tons during the two periods, and the NH₃-N between them are 263.200 tons and 237.625 tons, respectively. The TN and TP have shown a decreasing pattern in the two periods. We estimate that TN pollutant discharge decreased from 3221.300 tons to 2908.286 tons, and TP reduced from 1422.500 tons to 1284.275 tons. Therefore, this evidence clarifies that the COF program has effectively controlled the soil and water pollution in DHW.

4.2. Socio-Economic Consequences of the COF Project

In terms of consumption activities (

Table 1. Average changes in expenditure portfolio (yuan per household) between COF participants and nonparticipants.

Consumption Activities in 2010 (before COF)	COF Participating Households	Nonparticipating Households	Simple Difference	Difference-in-Difference (ZDiD)	Difference-in-Difference with Matching (ZDiDM)
Consumption activities in 2010 (before COF)	A	B	A-B
Education, yuan/hh	4.401 ± 7.537	5.261 ± 11,562	−860
Natural resources
Wood, kg/hh	3.461 ± 4.282	5.284 ± 8.009	−1,823
LPG, yuan/hh	449 ± 615	318 ± 419	131
Cash gift	3.075 ± 4.127	3.438 ± 4.451	−363
Material assets
Automobile	0.03 ± 0.18	0.06 ± 0.24	−0.03
Motorcycle	0.59 ± 0.61	0.56 ± 0.54	0.03
Television	0.78 ± 0.57	0.82 ± 0.65	−0.04
Refrigerator	0.47 ± 0.52	0.60 ± 0.55	−0.12
Washing machine	0.47 ± 0.53	0.55 ± 0.55	−0.08
Consumption activities in 2015 (after COF)	C	D	(C-D)	(C-D)-(A-B)
Education, yuan/hh	7.117 ± 10,933	9.612 ± 17,457	−2.495	−1.635 (−1.775) *	−830.417 (−0.86)
Natural resources
Wood, kg/hh	1.809 ± 2974	1.457 ± 2.860	352	2.175 (−1.901) *	1.138.021 (−1.29)
LPG, yuan/hh	640 ± 678	672 ± 1.046	−31	−162 (0.831)	−242.708 (1.978) **
Cash gift	4.309 ± 5.414	5.368 ± 7.836	−1.059	−696 (−1.585)	−618.75 (−0.227)
Material assets
Automobile	0.12 ± 0.34	0.18 ± 0.41	−0.06	−0.03 (−1.824) *	−0.021 (−1.134)
Motorcycle	1.02 ± 0.60	1.03 ± 0.57	−0.03	−0.03 (0.163)	−0.146 (2.205) **
Television	1.03 ± 0.46	1.09 ± 0.51	−0.04	−0.02 (−0.966)	−0.146 (1.979) **
Refrigerator	0.85 ± 0.40	0.95 ± 0.35	−0.09	0.03 (−2.783) ***	−0.083 (−1.205)
Washing machine	0.82 ± 0.46	0.87 ± 0.45	−0.03	0.05 (−1.624)	−0.063 (−0.727)

Notes: For DIDM results, t-stats are shown in parentheses; *, **, and *** denote the differences are significant at p < 0.1, p < 0.05, and p < 0.01, respectively. LPG, liquefied petroleum gas.

), participants and nonparticipants increased their education expenses. They have both reduced their fuelwood usage and increased the liquefied petroleum gas. Hunting/logging of forests as cooking wood has been forbidden in China; therefore, the farmers have decreased the fuelwood use, and liquefied petroleum gas has become the primary energy consumption source. Meanwhile, they have increased the material assets, such as washing machines and automobiles during the two periods. When using the matching estimator method, they decreased their material assets because of the COF implementation. The findings indicate that COF participants’ orchard income decreased dramatically compared to nonparticipants, but off-farm income increased (

Table 2. Average changes in income portfolio (yuan per household) between COF participants and nonparticipants.

Income Sources	COF Participating Households	Nonparticipating Households	Simple Difference	Difference-in Difference (ZDID)	Difference-in Difference with Matching (ZDIDM)
Income sources in 2010 (before COF)	A	B	A-B
All income	85,283 ± 65,972	93,074 ± 122,093	−7.791
Orchard income	72,921 ± 60,218	64,806 ± 88,634	8.115
Other agricultural income	3.190 ± 2.394	1.680 ± 9.881	1.510
Off-farm income	9.132 ± 27,683	26,588 ± 69,117	−17,456
Income sources in 2015 (after COF)	C	D	C-D	(C-D)-(A-B)
All income	69,119 ± 77,411	120,616 ± 127,675	−51,497	−43,706 (−3.886) ***	−46,732(−3.724) ***
Orchard income	12,721 ± 28,201	61,097 ± 70,608	−48,376	−56,491 (−3.769) ***	−56,413(−3.345) ***
Other agricultural income	9.210 ± 5.872	3.295 ± 20,044	5.915	4,405 (1.381)	581(0.183)
Off-farm income	47,036 ± 65,384	56,223 ± 98,436	−9.187	8.269 (−2.764) **	9.100(−1.188)

Notes: For DIDM results, t-stats are shown in parentheses; *, **, and *** denote the differences are significant at p < 0.1, p < 0.05, and p < 0.01, respectively. The results demonstrate that COF participants’ orchard income decreased comparing with nonparticipants over the study period, but the nonparticipants’ income increased.

). Participating households decreased their household income by 19%, while nonparticipating households increased their household income by 23% during the two periods.

We evaluated the net income for cultivating orchards, and the ecological forest is around 64,6296 yuan per ha and −1167 yuan per ha, respectively (

Table 3. Cost, revenue, and net benefit of an orchard and ecological forest in yuan per hectare (compares orchard in 2010 and Cunninghamia in 2015).

Category		Orchard (Mean ± SD)	Ecological Forest	Opportunity Cost
Cost	Seeding	2.146 ± 1.247	950 ± 312
	Compound fertilizer	8.358 ± 4.003	325 ± 177
	Chemical fertilizer	10,677 ± 7.946
	Organic fertilizer	2.719 ± 1.975
	Pesticide	10,220 ± 7.541
	Irrigation and collection	2.553 ± 2.783
	Labor	14,597 ± 12,300
	Total cost	46,572 ± 16,529	1.167 ± 351
Revenue	Gross earnings	111,201 ± 55,074	0
Net benefit		64,629	−1.167	65,796

Note: the currency amounts in 2015 was discounted to the value in 2010 based on the inflation information.

). Moreover, approximately 15% of the participating households investigated (n = 236) clearly expressed they would support the continuation of the COF program, showing the program’s margin benefits for the regional human welfare improvements and the program’s pitfalls (Figure 5). Additionally, for the COF program’s water quantity and quality benefits to persist, providers must ensure the land-use conversion will produce the amelioration. However, approximately 93% of the participating households reported that they would return to the growing orchard again if the COF program stops because the benefits of orchard planting are larger than other types in DHW. The question of the COF program as a long-term approach to tackle the water resource problem will appear.

5. Discussion

PES is an appeal to develop and implement market-like mechanisms to generate investment in natural infrastructure where no markets exist [16,36]. The foundation of PES understands the positive externalities of conservation activities, and the opportunity cost should be compensated through financial subsidies to provide incentives to change land-use practices [24]. PES projects were designed to achieve the two wins of ecosystem service improvements and ecological economics sustainability across the world. There is a growing concern about whether PES programs have achieved planned goals of simultaneously improving water quality and quantity for downstream areas and the objective of increasing the human welfare of local households. The PES scheme’s success depends on local background, PES implementation, and social-economic environment [24]. While others have paid much attention to the biophysical, technical, and economic aspects of PES [37], sustainability may be jeopardized if the socioeconomic and policy considerations are not included in the design for the PES. The COF project in DHW aims to conserve the water ecosystems facing the deterioration of water quality and quantity by cutting the navel orange to plant a natural forest. A deeper understanding of farmers’ livelihood changes and behavioral responses to COF policies would help the government adopt more reasonable policies to respond to the above changes.

5.1. Implications of the Livelihoods Changes and Environmental Consequences

Once the PES scheme has provided substantial ES, the ecosystem providers should be compensated by government subsidies. Their land-use behavior should continue to provide the ecosystem service for a long time. More importantly, the payment amount to the ecosystem providers, i.e., the upstream householders, should beyond the opportunity costs [9]. Otherwise, the providers have not enough incentives to insist on land-use behaviors. In other words, PES will not worsen the household’s livelihood and should evaluate the PES effects scientifically. Both PES schemes and human development policies should improve human well-being. Whether poor people participating in PES will benefit is crucial for the PES projects’ long-term sustainability [24,32]. To ensure the sustainability of PES, greater supports are urgent to alleviate participants’ income change resulting from participating in PES schemes. Therefore, a diversified funding source and adequate compensation are necessary.

The research found that participants have decreased household income due to the changed orchard area. Therefore, we afraid that the participant household will replant the orange navel tree again when the COF project ends (Figure 5). Participants and nonparticipants increased the material assets, reduced their fuelwood use, and increased the liquefied petroleum gas. Meanwhile, Orchard production is different from other agricultural activities in that it is the primary income source for local households. Decreased participants’ livelihood may offset some of the beneficial effects of the program. The participant household’s sustainable livelihood will not merely depend on the COF project and the local government should provide more job opportunities for the local farmers through the establishment of a sustainable green economy and environment-friendly industry, such as eco-tourism high-quality organic agricultural production base construction. The COF program involves ecosystem service, land-use change, and household livelihood change. Therefore, the COF is a socio-ecological system (SES). From the SES perspective, we should emphasize the household livelihood changes and sustainable livelihoods, the payment standard and time, production activities in future research. From the policy perspective, the COF project implementation should not focus only on water quality and quantity for downstream areas and ignore the household’s livelihood change. The win-win strategies between ecological restoration and household livelihood improvements are the central concern for COF schemes.

This area faces many ecological problems, including mining activities, the over-harvest of large areas of orchards, landscape homogenization, and rapid poultry raising. We are afraid the continuing COF program will have adverse effects on the goal of social development. Households participating in the COF program compared with nonparticipants decreased their utilization amount in terms of nutrient application. Sustainability hinges on the dynamic relationship between society and nature [50]. The COF program is a complex social-ecological system, which is the focus of sustainability science. Considering the spatial organization of landscapes, SESs thinking will become more salient for landscape sustainability [50]. The COF program isn’t equal to landscape sustainability; stakeholders’ livelihood changes should be incorporated into SESs to achieve landscape sustainability.

5.2. The Implication of Cost-Benefit Comparison

Many PES projects try to achieve the maximum economic benefits in improving ecosystem service outcomes [24]. PES program implementation has a unique local context. The knowledge about PES implementation’s local social, economic, and political context is instructed to achieve sustainable PES. Costs and benefits that accrue to the different groups include farmer’s opportunity costs (OC), payments costs (PC), and reduced pollutant treatment costs (TC). The OC for ES providers is the difference in net income between orchard plantation and ecological forest, 65,796 yuan per ha. For downstream beneficiaries, their cost includes the payment of 30,000 yuan per ha to the upstream ES providers. The COF program benefits include the value of reduced cost for COD, NH₃-N, TN, and TP treatment owing to the COF program implementation. The benefits for the four items are 3.4687 yuan/ha, 0.0459 yuan/ha, 647.4157 yuan/ha, 1562.26 yuan/ha. We assess the benefit to be appropriately 2177.19 yuan per ha [51,52,53,54,55,56]. The program’s benefits are 6,677.19 yuan per ha, and the program’s costs are 106,473 yuan per ha. Therefore, our results show that the COF program’s overall benefits are only 6.27% of the COF program’s total costs to beneficiaries (Figure S1). In aggregate, these costs are about 14.347 times the benefit of the COF total program. The ecological compensation from downstream beneficiaries is urgent to cover the households’ income reduction. Moreover, the environmental consequences in DHW were not attributed to the COF implementation because some natural factors, such as precipitation rhythm changes and temperature differences, may also contribute to the improvements of water quality, which benefits the downstream stakeholders. Moreover, PES schemes uptakes only a small proportion of the values given by natural ecosystems. Existing ecosystems’ values are often outside of the PES boundary.

The success of PES varies with the local context, policy environment, and PES design and its implementation [24]. Conservation is most likely to succeed when benefits outweigh costs for all relevant stakeholders. Policy decisions are often assessed through cost-benefit evaluation, which can help make ecosystem service research operational [51]. The temporal trend of cropland abandonment under China’s two PES programs, Conversion of Cropland to Forest Program (CCFP) and Ecological Welfare Forest Program (EWFP) in Tiantangzhai Township, Anhui Province, indicated the household received a large amount of cash compensation, and it is successful in terms of environmental protection [23]. PES cannot be considered a priori as the most cost-effective policy option to achieve ecological goals. PES’s institutional features are found to be correlated with more favorable livelihood impacts, such as motivated buyers and sellers, high payments, high degree of voluntary participation, low transaction costs, and better access to alternative income sources [6,9,30]. The agreed payments should be higher than the opportunity costs of ES suppliers and lower than the willingness to pay of ES users [15]. However, most payment standards are formulated based on water pollution treatment costs. Previous experiences with incentive-based approaches suggest it is unlikely a PES approach will always improve livelihoods, increase ES, and reduce costs simultaneously. The opportunity cost of ES providers in the COF program is higher than the household’s net income, and the income of households participating in COF indicated a decreased trend. The relatively higher costs compared with benefits resulting from participating in the COF project are a tremendous obstacle for the PES project’s success and have potential threats to local households’ enthusiasm. Moreover, COF is not the only way to control water pollution in the Dongjiang area. Integrated water pollution control measures should be implemented in the entire watersheds; for example, the closure of small pollution enterprises, the establishment of a water-quality monitoring network, and the newly-created policy of River Leader.

5.3. The Effects of Landscape Change Upon Food Safety

Food safety triggered by the reduction of farmland area due to rapid urbanization and industrialization is an essential issue for the central government. The State Council has decided to guarantee the arable land no less than 1.8 billion mu, indicating the importance of the arable land red line. The southeast of Jiangxi Province with the hilly and mountain area isn’t suitable for large scale grain plantation. Therefore, it is urgent to protect the limited arable land. The COF project implementation will inevitably pose significant effects upon regional food production safety. After the orchard garden has been affected by the policy and insect pests, some farmers started to plant orchard replantation in the paddy field to earn money and other fruit production in the previous flat paddy field, because the income from orchard plantation is the primary resource for the entire family’s living. This plantation change phenomenon will bring more revenue for the local farmers on the one hand; however, the cultivated land has been replaced by orchard garden and will pose potential adverse effects upon rice production and food safety. According to our investigation results, the paddy field area owned by each participant and nonparticipant is only 0.021 ha and 0.016 ha, respectively. Therefore, the landscape transformation in DHW will result in potential food safety problems as the paddy field area continues to decrease (Figure 6). We have paid enough attention to the land-use transition from agricultural land to urban and industrial land while caring little about the internal change among agricultural lands.

5.4. The Ecological Compensation between Downstream and Upstream Communities

The PES scheme is pushed forward by the shortage of ecosystem services, focusing on water quality, flood protection, and biodiversity [9]. Since many benefits have public goods properties, the PES scheme could be achieved via regulation or financial subsidies. Therefore, many PES schemes are mostly relying on transactions driven by regulation or government payments PES. The upstream communities that are paid to change their land-use activities are quickly confirmed. However, PES schemes’ subsidy is often inadequate for the ecosystem providers to alter their conservation activities. Moreover, many poor households’ sustainable livelihoods have a considerable influence on the ecosystem service, guiding PES schemes focusing on livelihoods and poverty reduction. The lower watershed of DHW should pay the upper reaches of the watershed for enhancing or maintaining water quality and quantity. The COF program implementation is based on household participation; however, the compensation standard is relatively low and will conduct orchard production again when the COF ends (Figure 5). The governance in terms of the law, institutions, bilateral agreements is necessary to achieve regulatory demand. Additionally, for the water quantity and quality benefits of the COF program to persist, providers must ensure the land-use conversion that will produce the amelioration. The results obtained from the research indicates that it could provide helpful information for further land-use conversion and ecological policies. It also suggests that the COF scheme is not a panacea that could solve all the water quality and quantity problems that downstream stakeholders are facing. The COF project is unsustainable and will not produce more positive effects if the downstream stakeholders do not provide enough financial support to the upstream stakeholders. Moreover, investing money does not ensure the provision of valuable ecosystem services [9]. At the time, PES should guarantee the financial subsidies are used most efficiently.

Although the COF program will not tackle the entire water resource problem in the China Great Bay Area, it shows the potential role in interprovincial water resource coordination to achieve the ecological and economic benefits by implementing environmental compensation. The results show that the labor force allocation has been changed after they participated in the COF project. Many examples focus on interprovincial coordination agreements, such as between Anhui province and Zhejiang province [23]. This kind of arrangement is very crucial to ensure the win-win of upstream and downstream areas. Guangdong-Hong Kong-Marco Greater Bay Area is becoming a vast metropolitan area and bay area with the largest economic aggregation, estimated to be 120 million in 2050. Due to the lack of water resources in the bay area, the local government seeks to deliver water from the Xijiang River to Shenzhen and Dongguan city. The program will invest about 35.4 billion yuan in alleviating the water scarcity in Dongjiang River.

6. Conclusions

PES primarily aims to change stakeholders’ behaviors specifically related to land-use management practices to enhance water resources provision positively. PES is believed to be probably the most informative innovation in natural conservation [57,58]. The socio-economic and biophysical context under which PES transactions occur is a critical determinant of program access and outcomes. It is crucial to assess and understand the magnitude, direct (positive or negative), and determinants of livelihood impacts of PES policies. Understanding the PES program’s socioeconomic and environmental effects can help design a good policy to generate ecosystem services and improve human well-being. Generally speaking, the PES scheme has ecosystem externalities of human activities. The agent-based coupled infrastructure systems (CIS) framework is a useful approach to evaluate PES sustainability. The CIS’s upstream coupling may influence the conservation policy transformation and deliver goods and services via PES to downstream stakeholders. Due to the CIS’s cascade effects, the CIS’s downstream coupling should influence ES beneficiaries, and the upstream coupling should affect ES providers’ actions. Therefore, the CIS plays a vital role in connecting upstream and downstream stakeholders through socio-environmental interactions. In this study, we conducted a difference-in-differences match (DIDm) estimation method to clarify the livelihood changes and corresponding environmental consequences through a field survey in DHW, a famous production base of navel orange in China. The programs have positively affected the watershed environment directly and indirectly via water quantity and water quality improvements. Despite the relatively short time, the program has already demonstrated substantial ecological and socioeconomic impacts. This program provides essential insights regarding opportunities and challenges in the development, implementation, and sustainability of similar ecosystem service payment programs, at present and in the future, both inside China and around the world. Moreover, the CIS’s framework proposed in the research could also be applied in other similar areas. The ecological projects will be influencing the household livelihoods and environmental consequences, both for downstream and upstream stakeholders. The value obtained from the experiment will support the ecosystem service assessment across the world and propose a suggestion for decision-makers to adopt environmentally-friendly ecological policies to improve ecosystem services substantially. The PES scheme implementation could appear as useful lessons for other payment schemes in China and the rest of the world. The research has an important implication for targeting landscape sustainability in DHW. Meanwhile, the findings of this research show that the ecological policy should not focus on the improvements of biophysical elements, such as water quality and quantity alone, and ignore the sustainable livelihood at the same time. If the COF projects need to be continuously implemented, the positive and negative effects should be well assessed including environmental monitor projects and the high-level ecological compensation to the upstream communities to and guarantee the landscape sustainability in a long run. Moreover, the local government should carry out social system monitors in time to adjust the ongoing conservational land-use activities during the COF project implementation process.

However, the findings have several limitations. Householders’ livelihood changes may have driven by other factors, such as local industry developments, ecological conservation policies, and farmer’s habitats cultivated in a long history, that will be posing potential unobservable effects upon research conclusions. Second, the ecological improvements, indicated by COD and other chemical elements amounts, resulting from participating COF projects may attribute to the positive environmental policies adopted by the downstream stakeholders. In recent years, large-scale environmental protection measures downstream of the Pearl River Delta have been proposed and launched to tackle the increasing water and soil pollution. Therefore, the cost-benefit analysis, a crucial research perspective for landscape sustainability, would inevitably bring research bias and error owing to the complexity of SESs. The COF projects should be incorporated into regional comprehensive ecological policies practices to achieve the improvements of environmental quality rather than regarding the COF as a sole measure. Future research will add insights toward sustaining ecosystem service provisions, balancing tradeoffs, promoting regional and global cooperation, and achieving win-win-win at multiple scales.