Riparian vegetation restoration: Does social perception reflect ecological value?

Social‐ecological contexts are key to the success of ecological restoration projects. The ecological quality of restoration efforts, however, may not be fully evident to stakeholders, particularly if the desired aesthetic experience is not delivered. Aesthetically pleasing landscapes are more highly appreciated and tend to be better protected than less appealing landscapes, regardless of their ecological value. Positive public perception of restoration actions may therefore facilitate stakeholder involvement and catalyse recognition of ecological improvement. Here we aim to contrast aesthetical perception and ecological condition in headwater river reaches restored through passive ecological restoration in study areas in Portugal (Alentejo) and France (Normandy). We recorded structural and functional indicators of riparian vegetation to monitor the ecological condition of study sites along a passive restoration trajectory. Aesthetical perception indicators were assessed through stakeholder inquiries developed under a semantic differential approach. We analysed perception responses to changes in the riparian ecosystems resulting from passive ecological restoration across different geographical contexts and social groups. The analysed social groups comprised stakeholders (environmental managers and landowners) and university students (landscape architecture and geography students). Results indicate that (a) visual preferences often do not reflect changes in ecological condition, (b) perception of the restoration process is strongly context dependent, and (c) experience and cultural background affect perception of ecological condition across the different social groups analysed. Clear identification of relevant stakeholder groups (those interested in or directly affected by restoration), effective communication, and stakeholder engagement are therefore essential for assuring the success of river restoration projects.


| INTRODUCTION
The social dimension of riparian ecosystem restoration has gained increasing recognition, as ecological restoration cannot be properly undertaken without considering the socio-economic context of the ecosystem to be restored (Dufour, Rodríguez-González, & Laslier, 2019;Swart et al., 2018). Ecosystem recovery resulting from restoration efforts may not be properly perceived by the public and by stakeholders (those interested in or directly affected by restoration), as perception is frequently driven by aesthetical experience, rather than by the recognition of ecological quality (e.g., Hands & Brown, 2002;Purcell, Friedrich, & Resh, 2002). Aesthetically pleasing landscapes are more likely to be appreciated and protected than less appealing landscapes, regardless of their ecological value (Gobster, Nassauer, Daniel, & Fry, 2007). Conflicting aesthetic preferences and ecological objectives have been increasingly addressed within the socialecological dimensions of ecological restoration (Cottet, Piégay, & Bornette, 2013;Junker & Buchecker, 2008).
Research interest in social perceptions of the ecological condition of riparian systems has grown, particularly regarding aspects such as the perception of wood (i.e., woody debris) in rivers (Chin et al., 2014;Le Lay et al., 2008;Piégay et al., 2005), wetlands (Cottet et al., 2013), and invasive species (Cottet, Piola, Le Lay, Rouifed, & Rivière-Honegger, 2015). Research concerning social dimensions of restoration has addressed the effects of stream restoration on the perception of riverine ecological condition. The analysis of data from 26 river restoration measures undertaken in Germany highlights the important role of social perception in defining the success or failure of river restoration measures (Jähnig et al., 2011). In New Zealand, McCormick, Fisher, and Brierley (2015) discussed the extent of explicit management "cues" in the perception of the degree of achievement of restoration goals (i.e., how evaluations of naturalness are a function of the degree to which people perceive the landscape to be "cared for").
In an urban context, aquatic condition indicators were used to analyse social perception of water quality and pollution (Hong, Chang, & Chung, 2018). Other components of the river system, such as riparian vegetation, are particularly conflictive in terms of social perception due to potentially opposite understandings of functions and (dis)services (Dufour et al., 2019;Kondolf & Yang, 2008). Riparian vegetation can be valued for human well-being in terms of its contribution to biodiversity conservation and flood and pollution control (Naiman, Décamps, & McClain, 2005). However, it may also be negatively perceived, as the development of woody vegetation is frequently associated with land abandonment and neglected or unsafe terrain (Purcell et al., 2002). In addition to contrasting stakeholder interests, riparian vegetation processes are not equally valued everywhere, and regional complexity must be better understood in order to appropriately adjust restoration actions (for European examples, see Piégay, 2009, andHughes, Colston, &Mountford, 2005; for an example from New Zealand, see McCormick et al., 2015). Thus, although restoration actions concern both ecological and social dimensions, these dimen- Moreover, research on the social perception of restoration efforts across social groups has gained increasing legitimacy because, as Gobster and Hull (2000) and Kondolf and Yang (2008) stated, restoration projects are fundamentally a social phenomenon. Public acceptance and support may ultimately determine the success and sustainability of a project, and the restoration process can also be dedicated to the satisfaction of social needs. Various authors have conducted research on the relation between visual preferences and ecological quality in riverine environments. Preferences for "wild" versus "managed" riverine landscapes were investigated by Van den Berg and Koole (2006), and the authors concluded that open, managed river margins were aesthetically preferred over forested, nonmanaged, "wild" ones. Junker and Buchecker (2008) used photographic simulations to test aesthetic responses to river corridors under different postrestoration levels of naturalness and concluded that aesthetic preferences were positively related to ecomorphological indicators of high ecological quality and high naturalness. However, results from other studies indicate otherwise. For example, Zhao et al. (2017) suggested that the relationship between ecological quality and visual aesthetic preference is not linear due to intermediate stages in ecological quality evaluation. For example, dense vegetation cover can be perceived by the general public as "unsafe," thus lowering people's aesthetic appreciation of these sites.
There is ongoing debate about how different social groups perceive changes in riparian vegetation, particularly concerning changes resulting from passive restoration (PR). PR consists of removing human disturbances (e.g., fire, grazing, and abstraction of river water) in order to allow for natural or unassisted recovery (Holl & Aide, 2011). As PR involves minimal management intervention in an ecosystem, thereby allowing the conditions for natural succession to proceed, it may have ecological and economic advantages. PR has been applied to the removal of persistent disturbances such as grazing.
Browsing of seedlings by herbivores prevents riparian recruitment (Painter, Beschta, Larsen, & Ripple, 2018), whereas trampling provokes disturbance and erosion of soil and habitat degradation (Nomiya et al., 2003). In riparian ecosystems, herbivore exclusion may sometimes be all that is necessary to achieve restoration success, thus requiring a relatively small budget (Forget, Carreau, Coeur, & Bernez, 2013). However, PR approaches involve potential drawbacks in terms of perception. The longer recovery time typically required in PR can be perceived as project failure and, in the worst of cases, may lead to the premature termination of a project by a landowner who would like to see more rapid or visible results, because areas subject to PR are often perceived as unused land (Zahawi, Reid, & Holl, 2014). This is particularly relevant in headwaters, which represent 60-80% of the cumulative length of river networks across landscapes (Benda, Hassan, Church, & May, 2005;Brooks & Colburn, 2011), but which have received relatively little attention compared to larger rivers (but see Mallik, Newaz, Mackereth, & Shahi, 2011).
In this study, we aim to further investigate how people perceive scenic quality in headwater streams, to explore the relationships between such perceptions and ecological condition, and to examine how these perceptions change along a PR trajectory. Specifically, we

| Ecological condition
The two study areas are in the Alentejo region of Portugal and in Normandy, France. They are similar enough (see full descriptions below) to enable comparability, in that they are both headwaters and are both subject to similar environmental pressures and restoration approaches, but they have the differences required for the purposes of the study (i.e., they are located in different biogeographic regions).
The Portuguese study area is located in the Tagus river basin  Table 1) are classified as "rivers of the sedimentary deposits of Tagus and Sadotype S3" according to the The French study area is located in the Sélune river basin, Normandy, France ( Figure 1). The climate is oceanic, with a mean annual temperature of 11.6°C and an average rainfall of 760 mm/year (www.meteofrance.com). The study area's streams and rivers (see Table 1) are classified as very small streams of the Armorican Massif river typology developed according to the WFD criteria (Chandesris, Wasson, Pella, Sauquet, & Mengin, 2006). Sampling was conducted in the Oir River, a headwater stream with permanent flow and a drainage area of 84 km 2 . Its landscape has long been shaped by agricultural activities (which represent 94% of the landcover at watershed scale), leading to a heterogeneous landscape including crops and pastures surrounded by hedgerows and streams. Field data in the Portuguese and French study areas were collected by means of floristic inventories conducted in 15 × 1 m plots set up between the stream and the fence, in which all vascular plant species present were identified to species level (based on Castroviejo et al., 1986, and Stace, 2010. In the French study area, 36 permanent plots equally distributed along four sectors of the stream (four different landowners) were established. These were sampled in 2004 (NR),

(ST), and 2014 (LT).
In the Portuguese study area, a space-fortime substitution approach was used (Pickett, 1989) due to the absence of a long-term monitoring scheme. In this case, 13 plots were equally distributed across the three study sites (NR, ST, and LT sites) and were sampled in 2016.  . The plant functional traits and ecological features we used were life form (sensu Raunkiaer, 1934), life span, leaf persistence, dominant reproduction type, and ecological strategy (sensu Grime, 1977). These were retrieved from the BiolFlor Database (Klotz, Kühn, & Durka, 2002). Life form is a useful way to functionally classify species based on the location of perennating tissues in relation to ground level and is related to their capacity to persist under different environmental conditions and disturbances such as grazing (Whittaker, 1975). Life span is linked with population persistence and is associated with disturbance, whereby annual species characterize pioneer communities (Pérez-Harguindeguy et al., 2013). Leaf persistence relates to the nutrientuse strategy of a plant, thus providing an indirect index of important plant traits such as potential growth rate, nutrient-use efficiency, at different stages of the PR process. Analyses were performed using the FD package in R (Laliberté, Legendre, & Shipley, 2014). Comparisons among groups were performed by means of a nonparametric Kruskal-Wallis test (Sokal & Rohlf, 1995).

| Social perception: Semantic differential survey
Perceptions of landscape aesthetics and of ecological quality were investigated by means of a semantic differential (SD) survey (Osgood, 1964), in which colour photographs were used as visual stimuli. Landscape perception studies have largely used oblique terrestrial photographs (e.g., Clay & Daniel, 2000;Fyhri, Jacobsen, & Tømmervik, 2009;Natori & Chenoweth, 2008), as the representativity and reliability of this approach have been long established (Hull & Stewart, 1992;Nassauer, 1983;Shafer & Richards, 1974;Stamps, 1990;Stewart, Middleton, Downton, & Ely, 1984). The SD approach is designed to measure the connotative meaning of objects, events, and concepts, based on the use of bipolar scales defined by Osgood (1964). Each bipolar scale consists of a pair of opposing concepts (or antonyms), which are usually adjectives, such as "ugly-beautiful" or "continuous-discontinuous." The present study employed such scales to determine a person's subjective perception of (or affective reaction to) the qualities of headwater river reaches. Qualities were then evaluated using a continuum of five scores linking the opposites in each scale, which respondents mark to show how they score After performing a first trial to ensure linguistic clarity and avoid redundancy, three bipolar scales were eliminated, leaving 20 scales grouped in the following four domains: "A-Scenic Quality," "B-Ecological Structure and Function," "C-Naturalness and Management," and "D-Functions and Services for Society" (Table 2; see also the Data S1 in supporting information). Furthermore, special care was taken in meeting the requirements and recommendations underpinning the SD approach (i.e., ensure the existence of sets of bipolar scales covering the whole domain, check for linguistic and

| Ecological condition
The results of the ecological condition analysis (Figure 3)

| SD analysis
We did not observe significant differences in responses across gender, age, education, nationality, and country-of-residence groups.
However, we did find significant differences between geographical contexts of the analysed photo sets, stage in the restoration trajec-  Grime, 1997) and (d) community weighted means of R-ruderal plant species (classified according to Grime, 1997) along the passive restoration process (NR, nonrestored; ST, short-term restored; LT, long-term restored) in both countries (full bars, France; open bars, Portugal). CWM, community-level weighted mean able to distinguish between the restoration stages, notably for the "attractive for fauna" and "productive" bipolar scales. Moreover, for Portuguese sites, the Portuguese respondents (students and stakeholders) seem to perceive differences less than the French respondents; this was the case for the "sheltering" and "flood hinderer" bipolar scales, for example.

| DISCUSSION
The analysis of ecological condition revealed that PR was generally associated with increases in ecological complexity and diversity in the assessed riparian communities in both geographical contexts, with poorly structured riparian communities being found in the NR sites  Purcell et al. (2002) and subsequent comments by Kondolf and Yang (2008).
SD analysis showed that responses to the inquiry on aesthetic perceptions were significantly different among analysed groups and across the four perception domains. For certain domains, responses for some of the bipolar scales showed no differences among groups.
For example, the responses in the "A-Scenic Quality" domain seemed to be linked to a more immediate aesthetic impression and less related to previous experience or educational background. This could be interpreted as an "initial affective reaction" in Ulrich's model of affective response to a natural scene (Ulrich, 1983). Experience and cultural background seemed to influence perception of changes in ecological condition across the restoration process in the remaining three domains ("B-Ecological Structure and Function," "C-Naturalness and Management," and "D-Functions and Services for Society"). In these domains, stakeholder groups attributed different scores, presumably due to their closeness to and experience in the ecological context.
For example, French stakeholders clearly ranked the LT French photo sets as "stable," "diversified," "natural," "uncontaminated," and "sheltering" in contrast to French students (less experienced) and Portuguese respondents (more familiar with the Mediterranean context).
Furthermore, stakeholders (both French and Portuguese) generally exhibited a more distinct attitude concerning the capacity of the riparian community to protect against bank erosion (a clear management issue), particularly when considering the long-term stages of recovery.
According to Ulrich (1983), this could be interpreted as a response determined by a "post cognitive affective state," a less immediate stage in the perception process that is much more influenced by one's culture and experience ( Figure 5). Ulrich's model is also in line with the findings of Zajonc (1980), according to which the occurrence of affective discriminations (like-dislike ratings) in the human mind happen "sooner than" and "independently of" the cognition processes supporting human judgments based on experience.
When analysing the changes in ecological condition and visual preference together along the restoration trajectory, we verify that the ecological changes implied by PR in the Portuguese photo sets tended to be negatively perceived from an aesthetical perspective and were scored more highly on the "neglected," "unproductive," and sometimes "polluted" bipolar scales. This contrasts with the result for the French photo sets and with the observations of Cottet et al.
(2013) concerning aquatic wetlands. Although people's aesthetical preferences have provided good support for several river health indicators in other regions of the world, subtle cues of human modification were found to be associated with perceived higher environmental quality (McCormick et al., 2015). This is in line with our results, where visual "indicators of human intention to care" (Nassauer, 1995) were positively perceived. For example, the fact that the French long-term-restored photosets were better perceived may be related to the fact that they showed an apparently more managed vegetation cover with aligned trees and a low herbaceous layer, whereas the Portuguese LT site was characterized by a much denser and more shrubby vegetation. This is particularly relevant in a headwater context, where the geomorphological conditions of small rivers (Benda et al., 2005) might prevent attractive open views of the landscape, leading to lower appreciation by the public for reasons unrelated to the restoration process and due simply to intrinsic spatial features of headwaters.
Overall, our results confirm that aesthetical assessment is highly context dependent, and thus, the combination of ecological and social dimensions of restoration projects reveals the complexity of river restoration assessment. Moreover, recognizing ecological improvement is often dependent on technical experience; therefore, it can be perceived differently by different groups or by experts with different educational backgrounds. Misalignments between the recognition of improved ecological conditions and social acknowledgement of restoration interventions may lead to misunderstanding among managers, landowners, or other societal groups and may ultimately determine the success or failure of restoration (Jähnig et al., 2011). This highlights the need to monitor not only ecological improvement following river restoration but also to assess how restoration is perceived socially in order to promote effective engagement of stakeholders in the restoration process. Such assessment can be incorporated into the planning and design of the ecological restoration programme to better integrate positive social perception as an explicit key element of restoration.

FUNDING INFORMATION
The