Participatory landscape sustainability assessment: where do we stand? A systematic literature review

In line with inter- and transdisciplinary approaches promoted in Sustainability Science, Participatory Landscape Sustainability Assessments (PLSA) are developing at a rapid pace. PLSA approaches share with other assessments the aim of standardizing observations, while sharing with participatory processes place-based and context-specific viewpoints from diverse stakeholders. This literature review presents different PLSA approaches identified in studies, and argues that the lack of a coherent framework and poor substantive theorization can limit the development of PLSA research. The study involved a systematic literature review on a corpus of 425 publications, combining bibliographic mapping on the full corpus and a content analysis of a sub-corpus of 138 full texts. The review of the literature showed that (i) PLSA studies lie at the intersection of ecology, landscape planning and sociocultural approaches, (ii) PLSA indicators evaluate on average 4.7 categories of sustainability, but most are applied at a local level and provide a snapshot of a situation, (iii) stakeholders tend not to be involved in the choice of indicators (only 28.9% of studies) and even more rarely in assessment design (7.2%). When stakeholders are included, they are usually only asked to populate preidentified indicators (63.9%). (iv) Diverse viewpoints are taken into consideration mainly by using indicators (67.3%) rather than by promoting discussion (39.8%). Three types of PLSA study can be differentiated: the participation-oriented approach, the contributive approach (rooted in positivism) and the collaborative approach (rooted in constructivism). We advocate that future PLSA studies pay more attention to consistency between their objectives, the methods they employ, and the theoretical grounding they enlist. This might help to avoid confusion about different participatory approaches and to understand their respective contributions to Landscape Sustainability Science.


Introduction
Public participation in decision-making is widely promoted as a cornerstone in landscape planning and is seen as an important precursor to landscape sustainability (Roe 2013;Calderon and Butler 2019). Landscape sustainability can be defined as "the capacity of a landscape to consistently provide long-term, landscape-specific ecosystem services essential for maintaining and improving human well-being" (Wu 2013, p. 999). In a perspective of post-normal science, landscape sustainability is generally considered a wicked problem, in which facts are uncertain, values in dispute, stakes are high and decisions are urgent (Funtowicz and Ravetz 1993). In such a context, decisions made by experts reveal their limitations, while stakeholder participation is considered to be both more ethical and more effective (Funtowicz and Ravetz 1993).
Participatory landscape sustainability assessment (PLSA) refers to processes that involve non-academic stakeholders in the various steps of landscape sustainability assessment, such as the determination of the scope and objectives of the assessment, the choice of indicators, their baselines, projected scenarios and values, and the application of the resulting practices (Dale et al. 2019). From an ethical point of view, PLSA aims to empower communities and to satisfy people's right to participate in decision-making processes, in particular those who are usually absent from non-participatory assessment approaches (i.e. those with little socio-political or economic capital) (Roe 2013). It aims to recognize the knowledge and values of everyone who is involved with a given landscape (Roe 2013;Calderon and Butler 2019). From a pragmatic perspective, public participation -in particular, participatory assessment -is advocated because it is (i) grounded in local contexts, (ii) able to provide concrete information for both policymaking and development projects, and (iii) likely to produce assessment indicators that better fit the vision of local stakeholders (Reed et al. 2006;Sébastien et al. 2017).
A PLSA approach is one way to satisfy the objectives of sustainability science, which is defined as a problem-driven, solution-oriented science that aims for sustainable development and takes an interdisciplinary (i.e. collaboration between multiple scientific disciplines) and transdisciplinary approach (i.e. the participation of non-academic spheres in research) (Clark and Dickson 2003;Kates 2011;Brandt et al. 2013). This has led to a call for Landscape Sustainability Science (LSS), in which landscape scientists play a pivotal role in understanding sustainability issues, taking a multi-scale, spatially explicit, and inter-and transdisciplinary approach (Wu 2013). Rooted in landscape ecology, LSS has evolved to become increasingly holistic, transdisciplinary and normative (Zhou et al. 2019, p. 274).
Yet within LSS, one of the knowledge gaps is a set of standards, methods and tools to evaluate the current state and future trajectory of landscapes, in order to guide these toward a sustainable future (Wu 2021). In order to develop these, a key preliminary step is to choose appropriate principles, dimensions, criteria and indicators. Specifically, a principle is defined here as a fundamental rule or law, an ideal that serves as a basis for reasoning and action (Bueren and Blom 1997); in this case, the overall principle is landscape sustainability. The dimensions of this are narrower fields of the overarching principle, such as the "three pillars of sustainability" inspired by the Brundtland Report (economy, society and environment) (Wu 2013) or the five dimensions set out by Selman (2008) (environment, economy, society, politics and aesthetics). The criteria are intermediate points on which information can be gathered that allow an interpretable assessment (Bueren and Blom 1997). For example, Dale et al. (2019) break down the environmental dimension of landscape sustainability into six criteria: soil quality, water quality and quantity, greenhouse gas emissions, air quality, biodiversity and productivity. Lastly, indicators are qualitative, quantitative or descriptive variables that are periodically measured to monitor the criteria and their trends (Bueren and Blom 1997). For instance, three indicators can be used to measure soil quality: total organic carbon, total nitrogen and extractable phosphorus (Dale et al. 2019).
One difficulty with PLSA is that the very proposition can seem contradictory, as it treads a line between two divergent aims (Sébastien et al. 2017). On the one hand, assessments require a certain degree of standardization and normalization of measures and observations; on the other, a participatory approach must integrate a plurality of perspectives, acknowledging all stakeholders and viewpoints as relevant and legitimate, incorporating context-specific features that are, by definition, not generalizable (Sébastien et al. 2017;Díaz et al. 2018). As a result, tensions can arise at three steps in the assessment process: in the choice of scale, in the choice of indicators and in the inclusion of diverse viewpoints.
Regarding the first challenge, PLSA studies aim to take into account different scales, from local to global. Indicators are generally measured locally and are context-dependent, yet the assessment process aims at standardization to guarantee comparability between study sites (Sébastien et al. 2017). A landscape approach seems key to address these tensions between context-specificity and generalization. According to Antrop (2000), a landscape approach is holistic (considering that each element gains its significance because of its position and relationship with surrounding elements), relativistic (considering that landscape observation is subjective and depends on the characteristics of the observer) and dynamic (considering that the functioning of a landscape depends on its changing structure).
Focusing on the scale of the landscape allows data and results to be linked across several scales (Termorshuizen and Opdam 2009;Vallés-Planells et al. 2014;Englund et al. 2017;Wu 2021) as well as with social and ecological issues (Vialatte et al. 2019). From a governance point of view, a landscape scale also allows power issues to be addressed (Roe 2013) and stakeholders from the same inter-knowledge network or professional or geographic proximity to be brought together (Lepart and Marty 2013). This article will discuss the contributions of PLSA to the "scale debate" in landscape sustainability assessment.
Regarding the second challenge, PLSA must be based on a standardized assessment process to ensure scientific rigor, reproducibility and comparability, yet while incorporating diverse perspectives from all stakeholders. In terms of the choice of indicators, in theory, stakeholders should be involved in every step of a landscape sustainability assessment, from the definition of the overarching principle to the choice of the criteria to assess and the indicators to measure these criteria (Reed et al. 2006;Dale et al. 2019). However, in practice, many studies rely on a predefined list of indicators and involve stakeholders only at a late research stage (Sébastien et al. 2017). Similarly, the level of participation can vary, from providing information to empowerment (Brandt et al. 2013), or from contributing ideas to co-creation (Bonney et al. 2009). This article will discuss the various roles of stakeholders in PLSA and will explore the different theoretical approaches underpinning these differences.
The third challenge in PLSA is integrating a variety of viewpoints in a normalized assessment process (Sébastien et al. 2017). Landscape research has a long tradition of dealing with diverse representations and viewpoints. As the European Landscape Convention states (Council of Europe 2000, Ch.1, Art.1a), landscape is "an area, as perceived by people". Many concepts, such as local and indigenous knowledge, identities or traditional values, are required to capture this diversity (Roe 2013). In an assessment process, this diversity of viewpoints can be included through specific indicators (aesthetics, for example) or through discussion (to determine the contribution of each indicator to landscape sustainability). This article will look at the strengths and limitations of both approaches.
There have been a number of literature reviews on landscape sustainability (Huang et al. 2015;Cohen 2017;Zhou et al. 2019;Gibbes et al. 2020;Liao et al. 2020;Oudes and Stremke 2020;Sheikhnejad and Yigitcanlar 2020;Medeiros et al. 2021;Puskás et al. 2021;Yang et al. 2021), but none of these specifically addresses PLSA. This literature review seeks to fill this gap, with the aim of helping scientists and practitioners better understand the different approaches that coexist in PLSA, in particular with regard to how stakeholders actually participate in landscape sustainability assessment, and to the different theoretical approaches underlying PLSA research. This article aims to contribute to the development of participatory landscape planning from a substantive standpoint, in line with Calderon and Butler (2019).
The article presents the five key results of the literature review, summarizing (i) the variety of scientific approaches and methods, (ii) the use of indicators, which are mostly localized and provide a snapshot of a situation, (iii) the varying level of participation; (iv) the varying role of stakeholders in the coproduction of indicators, and (v) the inclusion of diverse viewpoints. The article also discusses the contribution of participatory studies in multiplescale landscape sustainability assessments and the importance for researchers to be explicit about their theoretical approach.

Literature search
To ensure a transparent, reproducible and robust literature review, we followed ad-hoc international standards (Pullin and Stewart 2006;Haddaway et al. 2018). We first built a search string that aimed to identify literature on PLSA (Fig. 1). The search string included four sets of keywords that respectively aimed at identifying publications dealing with (i) participation, (ii) sustainability, (iii) landscape and (iv) assessment. The search string included synonyms (for example, *valu* instead of assessment), as well as lexical fields (for example, ecosystem/landscape service and nature('s) contribution to people, which are concepts that have been key in environmental assessments since the 2000s (Jacobs et al. 2016) (Table S1).
No restriction on language or publication year was applied, but the search was limited to articles, journals, book chapters and books. The search was conducted on Scopus on August 23, 2021, and returned 496 records. We chose to search Scopus rather than the Web of Science to allow larger journal coverage (Mongeon and Paul-Hus 2016); it should be noted that both databases are overrepresented in publications in English and also lean toward the Biomedical, Engineering and Natural Sciences to the detriment of Social Sciences and Arts and Humanities (Vera- Baceta et al. 2019). Another reason we chose Scopus is that it provides exhaustive metadata for each record, allowing in-depth bibliographic mapping.
In a second step, we screened all publication titles and abstracts to exclude non-relevant records. Specifically, studies were excluded if the word "landscape" was used as a metaphor (e.g. "commercial landscape") and/or if the word "participate" did not concern humans (e.g. ountries). After screening, we obtained a literature corpus of 425 records (Table S2) that we analyzed with scientometric techniques (see Sect. "Bibliographic mapping of the entire literature corpus"). We then randomly selected a literature subcorpus of 138 publications and performed a content analysis on the full texts of these publications (see Sect. "Content analysis of the sub-corpus").

Bibliographic mapping of the entire literature corpus
We used R statistical software (Team 2021) to analyze the evolution in the annual number of publications on PLSA, as well as VOSviewer software (van Eck and Waltman 2011) to analyze keyword cooccurrence in order to highlight the main topics (and their interrelation) addressed in the literature corpus (a thesaurus was used to merge synonyms in keywords and in journals; see Tables S3 and S4). We also conducted a co-citation analysis to evaluate the scientific approaches of the publications in the corpus. Cocitation analysis was also carried out at journal level to assess the scientific links between journals.

Content analysis of the sub-corpus
For the content analysis, we randomly selected 33% (N = 141) of the publications in the corpus, which resulted in a sub-corpus of 138 records (out of the 141 publications, 3 were unavailable to review) ( Fig. 1). This sampling effort was a trade-off between statistical robustness and representativeness, and time and human resource constraints.
We designed an analytical grid containing a total of 53 variables (Table S5). For each individual publication in the sub-corpus, we extracted and coded data on (i) the information in the study (1 categorical variable); (ii) the studied landscape(s) (2 categorical and 1 Boolean variable); (iii) the approaches and methods (9 Boolean variables); (iv) the indicators used (3 categorical, 14 Boolean, 1 quantitative variable); (v) the implementation of the participatory process (15 Fig. 1 Main steps in the identification of the corpus Boolean and 1 quantitative variable); (vi) the outputs of the participatory process (6 Boolean variables). One researcher filled in the grid for all 138 studies to avoid bias due to multiple operators.
Once completed, the grid was imported into R for quantitative analyses with the Tidyverse, ggplot and FactoMineR packages (Lê et al. 2008). To analyze publication diversity, we conducted two Multiple Correspondence Analyses (MCAs), followed by Hierarchical Clusterings on Principal Components (HCPCs) on the articles that included case studies (N = 127). The HCPCs identified clusters of publications based on the squared Euclidean distance associated with Ward's agglomerative method (Husson et al. 2017). The number of clusters highlighted in the results was automatically calculated as the one with the higher relative loss of inertia; we applied 1000 iterations for the k-means consolidation. The first MCA/HCPC was based on indicators characteristic of active variables (N = 18 variables), plus two supplementary variables to help for interpretation. The second MCA was based on active variables of methods, indicators, role of participation in the study, and the involvement of stakeholders (N = 20 variables), plus seven supplementary variables.

A field at the intersection of ecology, landscape planning and sociocultural approaches
The results show that the literature corpus on PLSA has been continually increasing since 1992, with an inflection during 2004-2006 (Fig. S1). Analyzing co-citations at journal level identified three main scientific approaches, with strong links between them (Fig. 2). First, the "ecology" approach (red cluster) was represented in conservation ecology journals (Conservation Biology, Biological Conservation, Biodiversity and Conservation), sometimes taking an applied perspective (Environmental Management, Ecology Application). Second, the "landscape planning" approach (blue cluster) was found in landscape governance and planning journals (Landscape and Urban Planning, Land Use Policy, Journal of Environmental Management), and some journals with an integrated approach (Geoforum, Sustainability). Lastly, the "socio-ecological" approach (green cluster) was represented in socio-ecological system journals (Ecology and Society, Ecosystem Services) and in disciplines such The analysis of author keyword co-occurrence resulted in a further four clusters: biophysical dimension of landscape, cultural approach to landscape, political dimension of landscape, and landscape as a development tool (Fig. S2).
In the sub-corpus of 138 publications, 127 publications were case studies, 10 were conceptual or methodological studies and one was a literature review. The case studies were mainly located in Europe and central Asia (N = 78), followed by East Asia and the Pacific (N = 22) and North America (N = 19), and 8 were in other regions. Studied landscapes were mainly rural (N = 66), followed by urban landscapes (N = 25) and coastal landscape (N = 10), while 26 studies focused on multiple landscapes. Finally, 78 studies were not protected areas, while 49 studies focused on a landscape with a protected area.
The content analysis revealed three ways in which authors considered participation in their study. In 15% (N = 21 studies) of the sub-corpus, participation was not concretely enlisted but only mentioned in the introduction or discussion. About 70% of the sub-corpus (N = 97 studies) concretely used participatory methods, while 14% (N = 20 studies) considered participation as a research topic (for example, analyzing public participation in urban planning), without initiating a participatory process.
Of studies that included participation, a variety of participatory methods was observed (Fig. 3). Qualitative methods were the most frequent (N = 54), including, for example, semi-structured interviews and field observations (Vallet et al. 2019) or photo elicitation (Altaba Tena and García-Esparza 2019). These were closely followed by collective assessments (N = 48) relying on focus groups (Chakraborty et al. 2018) or collective workshops (Sapkota et al. 2021), and quantitative methods (N = 47) based, for instance, on surveys and household-level socio-economic assessments (Sharma et al. 2016). Scenario-based approaches (N = 17) such as scenario analysis (Bohnet et al. 2011) or scenario planning (Karjala and Dewhurst 2003), and modeling approaches (N = 18) such as agent-based modeling and Bayesian networks (Shaaban et al. 2021) were the least frequent. Less than a third of the studies used spatially explicit approaches (N = 45) such as remote sensing and geographic information science analysis (Kivinen 2015) or participatory mapping (Bélisle et al. 2021).
Many studies combined several methods (Fig. 3): 82 case studies (64.6%) used more than one method, including 50 (39.4%) that used three methods or more. Collective and spatially-explicit methods were more frequently combined with qualitative than quantitative evaluation methods. For example, Baumann et al. (2020) used a "tricot approach" that combined participatory sketch mapping, transect walks and semi-structured interviews, while Abeywardana et al. (2019) combined semi-structured interviews, field observations and participatory mapping.
Landscape sustainability indicators are mostly localized and short-term Short-term, or even snapshot, local-level studies were overrepresented in the literature (Fig. 4a). Results showed that 54% of the sub-corpus articles focused on a local scale, while only 5% used multiple-scale approaches. Yet some local studies adopted a multiple-scale approach by taking into account several local levels: e.g. from architectural level to public space and urban planning level (Wulff Barreiro and Brito Gonzalez 2020) or from farm to municipal level (Gullino et al. 2018). Regarding temporal patterns, Fig. 3 Network of the methods used in the sub-corpus studies (N = 138). Circle sizes are proportional to the number of times the method was used. The thickness of the links between two circles is proportional to the co-occurrence of two methods in the same study. Colors were chosen arbitrarily; they have only an illustrative function we found that 46% of the sub-corpus articles provided a snapshot of a given time; no study linked several temporalities.
The most frequently-used assessment indicators were related to biodiversity, agricultural outputs and land use, as well as physical and psychological experiences (Fig. 4b). In most studies, several categories of indicators were used, with an average of 4.7 (± 2.6) categories per study. The MCA and HCPC based on the characteristics of assessment indicators allowed us to identify three clusters of contrasting characteristics (Fig. S3, and Table S6 for full cluster description): the first cluster (N = 53 studies) included studies that mostly relied on qualitative indicators. The second cluster (N = 24) included studies that used indicators related to air (54.2% of the cluster while this indicator was present in only 11% of the subcorpus), water (87.5% in the cluster vs 51.2% in the sub-corpus), biodiversity (91.7% vs 56.7%) and physical and psychological experiences (83.3% vs 49.6%). Urban landscapes were overrepresented in this cluster (41.7% vs 19.7%). The third cluster (N = 50) included studies that used indicators of land use (88% of the cluster vs 48% of the sub-corpus), agricultural outputs (94% vs 58.2%), soil (70% vs 38.6%) and biodiversity (84% vs. 56.7%). We observed continuity rather than a sharp separation between the three clusters, suggesting a gradient between them.

A varying level of participation
We found three levels of participation, according to (i) how stakeholders are involved or not in the assessment process and (ii) at what stage they are involved in the process (Fig. 5).
The MCA and HCPC analyses revealed three clusters, corresponding to three levels of participation or coproduction (Fig. 5, see Table S7 for cluster description). In the first cluster (51 items), the coproduction type is termed "co-data:" stakeholders only provided information for indicators (86.3% of the studies, while this modality was present in 48.8% of the sub-corpus). This cluster is also characterized by the use of quantitative methods (74.5% in the cluster vs 36.2% of the sub-corpus), the use of subjectivity-based indicators (80.4% vs 55.1%) and the absence of discussion (96.1% vs 74%). The second cluster (45 items) includes two types of coproduction: "co-indicators," in which the indicators were chosen with stakeholders (42.2% of the cluster vs 20.5% of the sub-corpus), and "co-design," in which stakeholders were involved in the design of the assessment (13.3% vs 5.5%). In this cluster, the specificities are the use of collective methods (77.8% vs 35.4%), scenario-based methods (33.3% vs 13.4%), qualitative methods (60.0% vs 41.0%) and discussion (64.4% vs 26.0%). The third cluster (31 items) consists of non-participatory studies (96.8% vs 24.4%).
The following section discusses two variables of the level of participation: the stage at which stakeholders are involved and the inclusion of diverse viewpoints.

Coproduced indicators are a minority
Our analysis revealed that only a minority of studies involved stakeholders in the PLSA process at an early stage. Of the 97 studies that involved stakeholders, three types of research practices were identified ( Table 1).
The first type of coproduction ("co-data") was used in 62 studies. In these studies, stakeholders were solely "informants:" they populated values for indicators, but were not involved in the choice of the indicators themselves nor in the design process. These studies were dominated by quantitative approaches, with participants asked to provide data to calculate preestablished criteria and indicators. In general, the selected criteria were derived from major international frameworks or standardized methods, such as the ecosystem services framework (Boafo et al. 2014 (Fischer 2005).
The second type of coproduction ("co-indicators") was used in 29 studies. In these studies, stakeholders were both "coproducers of indicators" (they b Screeplot of the eigenvalues. c Dendrogram of the resulting HCPC with colors delimiting the three identified clusters: the "co-data" type (cluster 1), the "co-indicators" and "co-design" types (cluster 2), and the non-participatory studies (cluster 3) participated in the choice of indicators) and "informants." Some of these studies compared stakeholders' representations and standardized assessment frameworks, such as the Montréal Process Criteria and Indicators (Bayne et al. 2015) or Canadian or US national frameworks of criteria and indicators (Karjala and Dewhurst 2003). Other studies enriched existing frameworks, such as the ecosystem service framework (Bélisle et al. 2021), with stakeholders' representations. Some studies combined different frameworks, such as a study by Bravo-Monroy (2021) that used both the Lefebvre social theory and the Nature's Contribution to People concept to understand farmers' perceptions of space in Colombia. The third type of coproduction ("co-design") was used in seven studies that involved stakeholders as both "coproducers of indicators" and "co-designers of the assessment process". Some of these initiatives were led by authorities at the national level, as in the study by Pistorius et al. (2017) on forest landscape restoration in Ethiopia, while others were led by local people, as in the study by Metallinou (2020) on prescribed fires in Norway. In these studies, researchers sought to accompany, rather than lead, a process driven by local or national stakeholders.

Two ways of representing diverse viewpoints
Two ways of including diverse viewpoints on landscape sustainability were identified in the sub-corpus studies: through indicators and through discussion.
In 52% (N = 73) of the sub-corpus studies, diverse viewpoints were only addressed with so-called "subjectivity-based indicators," encompassing "physical and psychological experiences," "identities" and "learning and inspiration." The use of subjectivitybased indicators was meant to achieve three main purposes. First, to allow the study of the intangible dimension of landscapes based on scientific concepts such as attachment and identity (Bayne et al. 2015), sense of place (Urquhart and Acott 2014) or forest feeling (Korpilo et al. 2018). Second, to approach the intangible dimension of landscapes through local concepts such as "resourcing" held by First Nations in boreal Quebec (Bélisle et al. 2021). Third, to jointly investigate the tangible and intangible dimensions of landscapes in a materialist perspective. For instance, Tekken et al. (2017) developed a "portfolio of indicators" showing a negative correlation between the cultural ecosystem services provided by rice systems and land use intensity; Dobbie (2013) identified correlations between perception of aesthetics and the presence of landscape elements; and Huang and Sherk (2014) compared sustainability performance and visual preference of landscape elements.
In 25% of the sub-corpus (N = 35 studies), diverse viewpoints on landscapes were only addressed with discussion: for example, during collective workshops or focus groups. For example, Bohnet et al. (2011) developed a landscape toolkit through which they proposed three scenarios to stakeholders based on biophysical data and used the toolkit to guide discussions between stakeholders on the trade-offs between the scenarios and on priorities for the area. Similarly, Moraine et al. (2016) implemented participatory modeling with farmers and technical advisors to design integrated agriculture-livestock systems. Other initiatives were conducted with similar approaches, using serious games (Speelman et al. 2014), participatory scenarios (König et al. 2015;Karner et al. 2019), or multi-criteria decision processes (Terêncio et al. 2021). These initiatives show that even when the materiality of the landscape is approached with biophysical indicators, the link between the value of the indicators and sustainability needs to be discussed.
It is noteworthy that 20 studies (14% of the subcorpus) used both subjectivity-based indicators and discussion, whereas in 51 studies (36%), the issue of the divergence or convergence of viewpoints on landscape sustainability was not addressed.

Multiple ways to foster participation
Our bibliographic mapping showed that PLSA studies are cross-disciplinary and are conducted by diverse scientific communities, mainly in the fields of ecology, landscape planning and socio-ecology. This result is in line with Wu (2006), who identified two traditions in landscape ecology: a European tradition characterized by a society-centered holistic view focused on solution-driven research, and a North American tradition dominated by a bioecology-centered spatial view focused on question-driven studies.
Our results, in line with the typology proposed in Bonney et al. (2009), evidenced a gradient of participation in PLSA. One type of study (first cluster of the MCA; Fig. 5) took a contributive approach in which stakeholders were involved only to provide information about indicators that they did not choose. Viewpoint diversity was addressed through indicators, and the link between measured indicators and sustainability was not discussed. The second type of study (second cluster in the MCA) took a collaborative approach, in which stakeholders both chose and informed indicators. Qualitative and collective methods offered room for coproduction between researchers and stakeholders, to determine and discuss indicators and their measurement. The last type of study (third cluster) took a participation-oriented approach that only considered participation as an element of context, a perspective and/or a research topic.
In the corpus we identified, the authors who claimed to use participatory approaches differed in the methods used, the role they attributed to stakeholders, and the way they accounted for subjectivity (or not). While this variety of approaches and methods enriches the scientific field of PLSA, there is a risk that the lack of a consistent framework and clarity around the term "participation" can create confusion. To avoid this, we argue that it is important to be explicit about the theoretical approach underpinning the participatory process. To this end, the following section focuses on the differences between the contributive approach and the collaborative approach, since the participation-oriented approach does not actually use participatory methods.

Two theoretical approaches underpinning participation in landscape assessment
The contributive and the collaborative approaches differ in their methods, the indicators they use, the way they involve stakeholders and their diverse viewpoints, as well as in their theoretical grounding (Table 2). We argue that the contributive approach is rooted in positivism, whereas the collaborative approach is rooted in constructivism. Both approaches have strengths and weaknesses. We advocate for more clarity on the theoretical approach taken by PLSA studies, so researchers can better align these with their objectives, and ultimately their methods and interactions with stakeholders.
About 40% of the sub-corpus (51 studies) dealt with viewpoint diversity through specific indicators, such as aesthetics or experiences, using quantitative methods. We argue that this approach is rooted in positivism, as theorized by Comte (1798-1857), which seeks to generate verifiable certainties with the help of empiricism (Kühne 2019). In this perspective, landscapes are understood as material objects with a set of observable, measurable and countable quantities, or composed of different analyzable "layers" that can be unpacked by disciplines such as ecology, geographic information science and geography, but also psychological landscape research, which analyzes landscape perception in a dissecting way (Kühne 2019). This approach generally goes hand in hand with consequentialism, which is based on the idea that a good decision is one that best improves collective well-being by optimizing the trade-offs between relevant indicators and affected agents (O'Neill et al. 2008). For studies rooted in a positivist and consequentialist perspective, it is perfectly coherent to rely on a contributive approach to participation, i.e. to enlist stakeholders only to provide preidentified indicators. As such, these studies effectively set aside the normative issues associated with the choice of indicators and rely on indicators defined by scientists as tools able to objectivize a reality independent from human subjectivity. The use of a standardized framework allows comparability between case studies, while context-specificity is reflected by indicators populated by stakeholders (Sébastien et al. 2017). In Table 2 Two PLSA approaches, characterized by their methods and their theoretical underpinning Participation approach (in reference to Bonney et al. 2009) Contributive Collaborative Theoretical approach (in reference to Kühne 2019) Positivist Constructivist Mode of coproduction (in reference to Chambers et al. 2021) Researching solutions Navigating differences Methods Quantitative Collective Scenario-based Coproduction type Co-data Co-design and co-indicators Integration of diverse viewpoints Subjectivity-based indicators Discussion reference to Chambers et al. (2021), who proposed a typology of coproduction approaches for sustainability, the contributive approach would fit in the "researching solutions" mode, in which researchers and decisionmakers employ "realist" methods to produce practical scientific knowledge in order to influence policy and actions. The contributive approach is instrumental and solution-oriented; and aims at produce practical scientific knowledge with the goal to influence policies and interventions (Chambers et al. 2021). From a pragmatic perspective, proposing an initial list of indicators to participants as a basis for discussion avoids fragmentation (Sébastien et al. 2017). However, this approach has several limitations. First, the indicators may not fit with stakeholders' representations, as highlighted in papers that point to discrepancies between stakeholders' representations and standardized indicators (Karjala and Dewhurst 2003;Balfour et al. 2021;Bélisle et al. 2021). Secondly, such an approach poorly addresses power asymmetries between stakeholders, and between stakeholders and researchers. Among others risks, decisionmakers can be more receptive to knowledge that helps rather than opposes their plans (Chambers et al. 2021). Another risk is that the assessment process can empower scientific knowledge while marginalizing the voice and knowledge system of other stakeholders excluded from the process who may be affected by the resulting recommendations (Chambers et al. 2021). Third, a contributive approach tends to reduce PLSA to a technical process, minimizing its normative and political significance. Excluding stakeholders from the process of choosing the indicators means that only a small group of people (i.e. the researchers or decisionmakers) choose which landscape characteristics or functions should be optimized. Ignoring the normative nature of assessment causes ethical problems by imposing a single vision of the desirable future for a landscape (Lepart and Marty 2013). Finally, this approach tends not to address issues of epistemic justice, generally adopting a hierarchical vision of multiple knowledge systems.
A contrasting approach adopts a greater degree of collaboration. We found this in 35% of the subcorpus (45 studies), which took into account diverse viewpoints through discussion, involved stakeholders in the choice of indicators, and used collective and scenario methods. We argue that such a collaborative approach is rooted in constructivism, in which landscapes are understood as individual or social constructs. Among the most significant authors taking a constructivist perspective are Schütz (1899-1959), Habermas (born 1929), Latour (1977-2022 and Callon (born 1945) (Kühne 2019). In contrast to positivism, constructivism considers a landscape a result of socially formed patterns of interpretation and evaluation, with an internal synthesis between material objects and symbolic meanings (Kühne 2019). Discussion and collective methods are preferred in a constructivist approach as decisions are not considered discrete events, but processes subject to constraints independent of the achievement of collective wellbeing. Consequently, landscape sustainability is not seen as a matter of optimizing indicators, but of establishing a deliberative process that ensures distributive justice and accounts for plural and incommensurable values (O'Neill et al. 2008). In other words, a constructivist approach to landscapes does not focus on "'what-is' questions, but [on] questions about who constructs the world and how, how world interpretations and world evaluations differ, and how they acquire social commitment" (Kühne 2019, p. 18). This explains why a constructivist approach involves stakeholders in the choice of indicators and engages a reflexive process on the strategic interest of indicators and on what "landscape sustainability" means in a specific context. For example, Karjala and Dewhurst (2003) compared the values attributed to the forest by the Tl'azt'en Nation with two national frameworks (Canadian Council of Forest Ministers; Local Unit Criteria and Indicators Development of the US Forest Service), showing that the criteria defined by the Tl'azt'en are highly interdependent, which is not the case in the national frameworks. They also identified locally relevant and acceptable sustainability thresholds to be incorporated into management plans (Karjala and Dewhurst 2003). Balfour et al. (2021) used a standardized framework (ecosystem services) as well as new indicators such as desirability (which includes animal welfare and land use intensity) and viability (financial, but also consumption behavior or reputational benefits). According to the Chambers et al. (2021) typology, the collaborative approach would fit in the "navigating differences" mode, which emphasizes processes of relating together, learning and empowering. In this mode, researchers hold less power and use facilitation techniques and boundary objects to explore conflicts and reframe perspectives in order to better address issues of recognitional justice.
The collaborative approach also has limitations. One of these is that coproducing an assessment process with stakeholders means high variability from one site to another, which reduces the possibility of comparing case studies. In some cases, high-level stakeholders can block the participatory process (Chambers et al. 2021). And while soliciting individual stakeholders to collect data as in the contributive approach is one thing, facilitating discussions between stakeholders with different representations, interests and levels of power is more complicated. When the researcher is a facilitator, the question of his/her attitude and the management of power asymmetries becomes central, which requires specific skills (Barnaud and van Paassen 2013).
The fact that the positivist approach appears dominant in PLSA literature is far from neutral regarding the issue of participatory processes in landscape assessment. Some argue that the spread of participatory assessment offers the promise of citizen empowerment and decreasing domination of participatory processes by experts (Sébastien et al. 2017). However, other authors argue that behind a show of empowerment, participatory processes in fact perpetuate expert dominance (Calderon and Butler 2019). Our results point to a paradox between the reinforcement of discourse on the importance of participation in landscape sustainability assessment and a predominance of expert views on the appropriate criteria and indicators to assess this. A central question is to what extent PLSA can contribute to increasing environmental justice, particularly in terms of distributive justice (a fair distribution of benefits from a landscape), epistemic justice (the recognition of multiple forms of knowledge within the assessment process), and procedural justice (the application of principles of justice in decision-making processes) (Loos et al. 2023). While distributive justice is frequently considered in PLSA studies, epistemic justice is addressed much more unevenly. We found that the majority of studies take a contributive approach in which the multiplicity of viewpoints on landscape sustainability struggle to be fully taken into account and in which the "experts" dominate. In terms of procedural justice, this concerns landscape planning rather than landscape evaluation: the aim would not only be to produce knowledge but to guide social change, which corresponds to "codesign/coproduction," an approach we found in only 7% of studies.

Relevance and limitations of the landscape scale for addressing sustainability issues
In PLSA, a landscape approach is seen as instrumental in better addressing sustainability issues as it is place-based and takes into account multiple scales (Sébastien et al. 2017;Wu 2021;). In the literature corpus we identified, and in the randomly selected sub-corpus, the majority were case studies, while conceptual and methodological studies were particularly poorly represented. In this regard, the PLSA corpus seems similar to that of LSS, which has been labeled a "place-based" scientific field (Wu 2013;Wu et al., 2021;Liao et al. 2020) that suffers from a lack of a coherent framework and substantive theoretical developments (Zhou et al. 2019). Such critiques have also been made of other research fields, in particular transdisciplinary research (Brandt et al. 2013) and landscape participation (Calderon and Butler 2019).
In theory, landscape has been advocated as being the most appropriate level to assess sustainability (Lepart and Marty 2013). In practice, in our analysis, the landscape level corresponds to the "local scale," which represented 53% of the sub-corpus. While this result confirms the importance of the landscape level in PLSA, it also reveals that a substantial share of this literature (almost half) does not concentrate on this level. This might be linked to landscape scientists advocating multiple-scale perspectives and linking several spatial as well as temporal levels (Wu 2021). This multiple-scale approach is put forward to justify the utility of participatory indicators, as these are seen as a way to connect context-specific dynamics (provided by local stakeholder knowledge and perceptions) with more general phenomena (from indicator standardization across case studies; Sébastien et al. 2017). Yet our results show that only 5% of studies combine several spatial levels, while none from our sub-corpus addressed multiple temporal scales. This result is in line with other recent literature reviews that point to the lack of multi-level and spatio-temporal approaches in ecological research on anthropized landscapes (Blanco et al. 2022).
Filling this gap by fostering multiple-scale PLSA would appear a key priority for future research given the valuable insights provided by the minority of studies that implemented this approach. For example, Karner et al. (2019) conducted a multiple-scale participatory process to compare land-sharing and land-sparing approaches in Europe. To do so, they developed a global scenario that they scaled down to five case studies. The global scenarios were designed by an expert advisory board, but the authors note that enlisting the expertise of local stakeholders would have been beneficial. Another study conducted by Marques et al. (2016) explored the role of stakeholders in the dissemination of the assessment outputs, and showed that farmer-to-farmer imitation is an important lever for the adoption of sustainable practices. These two examples show the interest of multiple-scale approaches to better address sustainability issues, and also illustrate that the challenges related to upscaling and downscaling can be better addressed through participatory processes. This seems to indicate that two key research challenges in LSS need to be tackled simultaneously: developing multiple-scale analyses and fostering stakeholder participation in research.

Conclusion
The PLSA corpus includes diverse fields, methods and indicators, which can be classified into three groups: participation-oriented studies, contributive studies and collaborative studies. The contributive approach has a theoretical underpinning rooted in positivism, and the collaborative approach is rooted in constructivism. Making these theoretical underpinnings explicit is valuable as scholars have pointed to the need to develop more theoretical and substantive studies in LSS (Zhou et al. 2019) and landscape politics (Calderon and Butler 2019).
Our review identified two pitfalls that should be avoided in future PLSA. One is a lack of clarity regarding the goals of PLSA and the posture of researchers. Despite an apparent consensus on broad concepts such as transdisciplinarity (Brandt et al. 2013) and landscape participation (Calderon and Butler 2019), our findings evidenced a large variety of methods and scientific approaches. Beyond just making a claim of participation, it is essential that researchers make their theoretical approach more explicit, in particular about the role they assign to scientists versus other stakeholders in the assessment process, and the way they handle (or not) their own subjectivity and their representations. This will ensure that their methods are attuned to their theoretical approach.
The second pitfall is the risk of falling into a socalled "tyranny of indicators," that is, confusing sustainability with its indicators (Brédif and Arnould 2004). A landscape assessment involves translating a complex reality into something simplified and measurable, therefore operating methodological as well as ideological choices (Brédif and Arnould 2004;Sébastien et al. 2017). Presenting assessment as technical and objective and making a claim for technoscientific rationality tends to obscure the political and normative dimensions (Sébastien et al. 2017). While we do not argue that all assessments should involve stakeholders in the choice of indicators, there should be better clarity about how, and by whom, indicators were selected. This way, rather than reducing the assessment of landscape sustainability to a series of technical measures decided solely by experts, it can play a more political role and engage with socio-environmental justice problems to contribute to landscape democracy.