The study area comprises the municipalities across the River Piedra (616 km²) in north-eastern Spain (Fig 2). The region is characterized by marked seasonal variability in the water flow. The upper part of the River Piedra (ca. 46 km) is dry for most of the year due to a combination of a semiarid climate and a calcareous substrate. The middle part of the river is permanent (ca. 30 km), as it receives groundwater discharge. River flow rates in the lower lands are usually altered with respect to natural rates depending on La Tranquera reservoir operations. This 78800000 m³ reservoir is regulated primarily for the benefit of irrigators from others basins.

The land cover of the floodplain is characterized by agricultural use (76%), including dryland cereal crops in the upper lands, irrigated cereal crops and poplar groves in the middle part, and fruit groves and orchards in the lower lands. Riparian forests (4.2%) are restricted to the upland gorges (usually dry) located between the municipalities of Aldehuela de Liestos and Embid, and to a private natural park, the Monasterio de Piedra, located in the municipality of Nuévalos. The park’s main attraction is the large number of waterfalls of the River Piedra, which contrast hugely with the semiarid surrounding landscape. The tourism generated by the park is the main economic driver of the area, and attracts tourists to other nearby amenities and activities (e.g., restaurants, lodges, trekking, mountain-biking, ornithology, fishing, kayaking). The population is dominated by elderly people and significantly more men than women, although this trend reverses during school holidays.

The SEM model explained a large proportion of the variation in the flow of ecosystem services, with a Relative Goodness of Fit of 0.858 (Table 4). A greater part of the variance was explained for environmental education (R²_a = 0.817), recreation (R²_a = 0.784), and nutrient regulation (R²_a = 0.488). Less variance (R²_a < 0.45) was explained for carbon sequestration, biological control, and raw materials.

The results of the SEM highlighted the fact that some ecosystem services were strongly related to others, while others were less influenced (Fig 4). Ecosystem properties (i.e., soil conditions and habitat quality) were key variables, given the significant influence they had on all ecosystem services to which they were related, except to freshwater supply. Soil conditions had a significant strong influence on nutrient regulation (β = 0.758). The largest influence of habitat quality was on environmental education (β = 0.904), recreation (β = 0.842), biological control (β = 0.615), and carbon sequestration (β = 0.524). Habitat quality also had significant but weak (β ≥ 0.1) influence on water quality and aesthetics, and a weak significant negative influence on nutrient regulation (β = -0.135). This latter could be explained by the different types of land uses included in the assessment, for which there might be opposite relationships. For instance, perennial forests have excellent quality (as assessed using the Riparian Quality Index [48]) but do not contribute much to nutrient regulation; conversely, cultivated land uses might have high concentrations of nutrients due to human inputs through fertilization. Intermediate services also had significant but weaker influence on final ecosystem services: biological control had a positive influence on raw materials (β = 0.379) and nutrient regulation had a negative influence on food production (β = -0.180), which can be explained by the fact that increasing food production can reduce nutrient regulation. Final services also displayed some interactions, namely, aesthetics had a weak significant positive influence on recreation (β = 0.153).

The contribution of each service is represented by arrow thickness in Fig 4 and highlights the main interactions maintaining the flow of ecosystem services. In our case study, supporting ecosystem properties (soil conditions and habitat quality) strongly influenced intermediate regulating services and cultural services, indicating that these are the key variables that mediate the flow of ecosystem services.

Four stakeholders groups were identified (see Table 3): primary sector, recreation sector, leisure and formal institutions. The identification of the ecosystem services linked to each stakeholder group was useful to detect key ecosystem services for each stakeholder group in terms of their use, the ability of stakeholders to manage each service, and power asymmetries derived from the management of ecosystem services (Fig 5). The primary sector (Group 1) was linked to most ecosystem services, either by using, co-producing, or impairing them; in addition, they were the main managers of two provisioning services (raw materials and food production). The recreation sector (Group 2) used and impaired water-related services, and used and co-produced carbon sequestration and cultural services, part of which they had great ability to manage. Leisure (Group 3) was linked to cultural and water-related services and to biological control. These three groups were also indirectly linked to habitat quality. The ecosystem services linked to formal institutions (Group 4) were used indirectly, except environmental education, which was co-produced by a section of this group (i.e., the government bodies and scientists), and used by the other section (i.e., the schools and universities). Further, this group was the main manager of habitat quality, water quality, and freshwater supply.

These results together with a general overview of the power relationships among stakeholders enabled us to classify the ecosystem services of this case study within a rival/non-rival and excludable/non-excludable matrix (Table 5, see S3 File for details).

We related stakeholders’ roles to the rival/excludable classification of ecosystem services by depicting each service in a diagram showing stakeholders’ use versus their ability to manage ecosystem services (Fig 6A) and stakeholders’ use versus their ability to impair ecosystem services (Fig 6B). The results on Fig 6A highlighted the effect of power relationships on access to ecosystem services and differentiated five types of clusters, which mostly corresponded to the previous stakeholder classification. Fig 6B provided complementary information especially useful to identify situations in which the same stakeholder group used and impaired the same ecosystem service.

We identified that formal institutions (Group 4) were the stakeholders with highest ability to manage and lowest use of ecosystem services (Fig 6A, top left corner). They managed a key supporting ecosystem property (habitat quality) and key intermediate regulating services (water quality) able to maintain the ecosystem services flow and, thereby, affect other stakeholders. Further, they managed final services, such as freshwater supply and recreation, without using them.

The recreation sector (Group 2) had the highest ability to manage and to use ecosystem services. This group contributed to produce cultural services by offering aesthetic enjoyment, recreation, and environmental education, exclusively managed some of these services, and also used them, benefiting from the tourism generated (Fig 6A, top right corner).

The leisure (Group 3) and primary sector (Group 1) were the stakeholders with the lowest or no ability to manage ecosystem services (Fig 6A, bottom right). Consequently, these stakeholders benefited from just the remaining non-excludable and non-rival ecosystem services (a part of cultural services, freshwater supply, water quality, and biological control).

Farmers from the primary sector (Group 1) were an intermediate cluster, as they had some opportunities to manage the services they used the most (nutrient regulation, soil conditions, and freshwater supply) through their farming practices (Fig 6A, right dotted line). However, this group impaired these same ecosystem services by overuse, driving non-rival ecosystem services to rival (Fig 6B, dotted line). Still in the second diagram, we observed that in the case of biological control, the use of the service did not directly imply degradation; rather this service was impaired by the farming practices used, which in turn may potentially affect other stakeholders. The impairers of habitat quality were leisure activities and the primary sector; however, they might not perceive it, as this service is not directly used but indirectly used through other services dependent on it. In addition, farmers were the unique producers of food and raw materials (Fig 6A, left dotted line). However, as these provisioning services are mostly exported outside the area, farmers just use them marginally (i.e., most of the production of such services is not on a self-consumption basis; rather, the income of these stakeholders mostly comes from the export of these goods).

Lastly, we identified a cluster that comprised those stakeholders having low use of ecosystem services and low ability to manage them. These were external to the ecosystem (e.g., the global food and raw materials markets and the carbon sequestration capacity of the atmosphere) (Fig 6A, bottom left).

The identification of keystone ecosystem services to maintain the flow of services (Fig 4), together with the identification of the stakeholder groups that used and managed them (Fig 6) highlighted the critical effect of power asymmetries on access to ecosystem services. In our case study, the strongest power was held by the formal institutions group through the management of keystone supporting ecosystem properties and intermediate regulating services, on which many other services depend or that are used by most stakeholder groups. For instance, habitat quality had the strongest effects on the majority of ecosystem services: environmental education, recreation, and nutrient regulation (Fig 4). Additionally, water-related services (water quality and freshwater supply) were the most conflicting services as they aggregated the largest number of beneficiaries and impairers (Fig 6B). As a consequence, the institutions group had the power to promote synergies and trade-offs between ecosystem services and the power to limit the use of those services they managed to specific stakeholders (e.g., regulations on freshwater supply and fishing permits), excluding others, and thus, creating potential social imbalances.

Additionally, the recreation sector had strong power as they managed and used the cultural ecosystem services driving the economy of the area. Finally, the primary sector had intermediate but still important influence on some ecosystem services. For instance, they were able to moderately manage the provisioning services on which their main income is based (i.e., food production and raw materials). Interestingly, the SEM revealed that, in our case study, these services were fairly disconnected from other intermediate services (Table 4 and Fig 4) because they were mainly dependent on human inputs (e.g., irrigation and fertilizers) rather than on ecosystem functioning. Additionally, some farming practices impaired critical ecosystem services (e.g., habitat quality, soil conditions, and nutrient regulation) that determined the integrity of intermediate and final services, thereby creating powerful feedback to the stakeholders using those intermediate and final services.

Integrating both ecological and social interactions along the flow of ecosystem services is key to understanding the likely asymmetries between stakeholders fostered by environmental management and to promoting sustainable management of ecosystem services [73]. Recent research has increasingly been addressing the flow of ecosystem services from production (supply-side) to use by society (demand-side) (e.g., [5,36,74,75]), and some have considered the implications of access to ecosystem services [76]. However, no framework has yet made explicit the existence of power relationships mediating both ecosystem services flows and stakeholder interactions. Specifically, recent research has pointed out the need to analyse the role of multiple stakeholder groups and their relationships with the provision, demand, and management of ecosystem services in order to contribute insights for sustainable management of ecosystem services [73]. In previous research, structural equation models have been used to test relationships between ecosystem properties and their effects on the provision of ecosystem services [77], and between ecosystem services and their effects on human well-being [78]. However, such analyses have not been previously connected to power relationships among stakeholders. Moreover, although Fisher et al. [79] discussed the importance of power related to poverty alleviation and ecosystem services, power relationships have still rarely been considered explicitly as a key factor determining asymmetries in the access to ecosystem services. Our study clearly demonstrates the relevance of power relationships in determining access to ecosystem services and its impact on the ecosystem services flow. Identifying and targeting such power relationships is essential for delineating environmental management policies while reducing trade-offs among ecosystem services [40] and thus, reducing social inequalities and conflicts.

The proposed framework was tested using our knowledge and data on the River Piedra case study and demonstrated its validity for uncovering three critical dimensions of ecosystem service actual supply: first, the dependence relationships among ecosystem services, and thus, the implications of their use and management; second, the formal power asymmetries between stakeholders; and third, the influence of such power relationships in the ability of stakeholders to access and to manage ecosystem services.

In addition, our work has been useful to test a methodology that identifies which stakeholders have the strongest power to mediate the flow of ecosystem services. These stakeholders were: (i) those controlling key ecosystem services (because they either can directly affect to other stakeholders, or can affect the capacity of the ecosystem to provide services); (ii) unique managers of a particular ecosystem service, and thus, able to control access and use of such service.

However, the application of such analyses in environmental management requires further work to specify and address informal and hidden power relationships between and within the stakeholder groups (e.g., social leadership, family ties, etc.). Other limitations for applying this study elsewhere are that it was time-consuming, as it required biological and social sampling; difficulties connecting biological sampling with social sampling to give insights into power relationships; and difficulties including the adjacent municipalities within the geographic boundaries (e.g., the river basin) but outside the targeted ecosystem (e.g. the river floodplain) in the power relationships analyses.

The classification of ecosystem services based on the concepts of excludability and rivalness (Table 5) concurred with most theoretical examples from Costanza [71] and Fisher et al. [3]. However, in our case study, we did not identify any examples of excludable but non-rival services, or non-excludable but rival, probably because management usually makes ecosystem services congestible (i.e., driving services from non-rival to rival). For instance, ecosystem services managed by the formal institutions group tended to fall into this category as this group can regulate the status of such services (e.g., policies to regulate water quality and waste water). Moreover, cultural services had two possible and opposite statuses: rival and excludable for activities performed on private sites or mediated by private companies (i.e., by the recreation sector), and non-rival and non-excludable, for those services enjoyed at open access sites. Hence, non-empowered stakeholders (i.e., the leisure group and the bulk of the primary sector) only had access to the remaining non-excludable and non-rival ecosystem services, and thus, are the most vulnerable stakeholders [72] at risk of being excluded from accessing the ecosystem services they need for their well-being.

These results contrast to other applications of this framework [80], indicating that the status of ecosystem services is highly context-dependent. In addition, classifying ecosystem services along the rival and excludable gradient enables accounting for the multiple possible statuses of an ecosystem service across land-use types and property rights. Indeed, the use of the land and access rights appeared as the critical factors determining the status of ecosystem services. Providing open access to lands and avoiding preventable dis-services (e.g., through conservation farming [81]) might change this classification. More importantly, focusing on managing the non-rival to rival movements of ecosystem services (i.e., the congestible services) would prevent their depletion.

As the proposed framework pinpoints, environmental management mediates the use of ecosystem services, and thus, their interactions [82]. Additionally, our results enabled us to distinguish how the management of each ecosystem service and power relationships among stakeholders may condition its status across the rival/excludable matrix. For instance, single-stakeholder management systems in which ecosystem services are used and managed by a single stakeholder group generated positive feedback. Such positive feedback had two opposite effects: they either reinforced the service, for example, recreational activities managed by the recreation sector attracted more recreational activities; or depleted the service, such as for instance, soil conditions and nutrient regulation managed by the primary sector were consumed at faster rates than normal recovery; or the overuse of water by leisure activities impaired their initially good status. Although single-stakeholder management systems should ideally lead to negative feedback or internal self-regulation, there is a high risk of eliciting positive feedback, in which the service is depleted by unregulated use, decreasing the capacity of the system to supply services in the long term.

In addition, we identified top-down management strategies, where management is made from the higher levels of governance – usually involving stakeholders external to the social-ecological system – to the local population. This was the case for the formal institutions group that managed habitat quality and recreation, and the recreation sector that managed cultural services. These management systems did not foster potential synergies among the ecosystem services supplied by the River Piedra floodplain, such as enhancing habitat quality and cultural services [29], and neither strengthened the communities’ governance of their resources. Rather, the population in this area is mostly dependent on external capital such as the European Common Agricultural Policy subsidies for farmers or the investments made by the main companies in the recreation sector. Top-down management systems often have low resilience [83,84] and can fail to resolve resource-users’ conflicts [13] or rather contribute to create new environmental conflicts [85]. However, examples of participatory bottom-up management systems such as decentralized forest management in Tanzania [86], coastal ecosystems in Kenya [87], and estuaries in South Africa [88] have proved to be important to complement existing top-down systems. In our case study and similar rural areas, such participatory systems could be implemented by local governments and mediated by bridging institutions at different organizational levels, and should be adapted to the cultural and geographical characteristics of each social-ecological system. In fact, bridging organizational levels through nesting institutions is one of the basic principles that have been identified for creating robust bottom-up management systems, together with deliberation and participation of interested stakeholders, institutional diversity (i.e. formal and informal institutions) [89] and knowledge leadership [90]. Consequently, encouraging knowledge exchange through participatory mechanisms is important to guarantee that multiple stakeholders regularly interact and discuss about their interests, needs and management of particular ecosystem services [91].