This study was conducted in the French Mediterranean region, in two departments with long-standing pastoral traditions: Hérault and Ardèche (Figure 1).

Hérault has a typical Mediterranean climate, with highly variable annual precipitation, concentrated in early spring and autumn. The landscape is characterised by shallow limestone soils and a mosaic of garrigue shrublands, woodlands, grasslands, and cultivated areas. Natural vegetation includes Mediterranean shrubs (e.g., Quercus ilex, Quercus coccifera, Cistus sp., Juniperus sp.), and herbaceous species such as Aphyllanthes monspeliensis and Brachypodium retusum. According to the 2020 agricultural census, the department supports 42,355 sheep and 4,515 goats, corresponding to 8.5 small ruminants per square kilometre (SR/km²).

Ardèche features an altered Mediterranean climate, with pronounced variation along an altitudinal gradient ranging from 200 to 1700 m a.s.l.: higher elevations receive up to 1800 mm rainfall per year, vs. 700 mm in the lowlands. Soils are mainly of granitic origin in the uplands, but volcanic and calcareous substrates are also present, creating a heterogeneous mosaic of pastoral habitats, including broom-dominated shrublands, deciduous forests, chestnut groves, wetlands, and upland grasslands. Common woody plant species and trees include Sorbus aria, Sorbus aucuparia, Cytisus scoparius, Genista scorpius, and the herbaceous layer is dominated by Festuca spp. The department supports 79,185 sheep and 32,952 goats, equivalent to 18.4 SR/km².

Local agricultural services provided us with a list of pastoral farms selected to capture a gradient of feeding systems in terms of reliance on rangelands and grazing management strategies. The farmers interviewed were selected according to four criteria: (i) significant use of semi-natural and wooded pastures (at least a few months per year, for part of the flock); (ii) livestock composed mainly of sheep or goats; (iii) interest in rangelands as a forage resource; and (iv) willingness to engage in a detailed discussion about rangeland utilization. The last two criteria were not measurable farm attributes, so they were evaluated based on the expertise of advisory services. The sample was not intended to be statistically representative of all pastoral systems in the study areas, but rather to capture a diversity of farming systems and decision-making processes across contrasting levels of reliance on rangelands and modes of grazing management. The farms included in the study were predominantly small-scale systems, with flock sizes generally below 500 animals, which reflects the structural characteristics of small-ruminant farming in the study regions. These systems are typically managed directly by farmers, who are closely involved in day-to-day grazing management.

Eighteen pastoral farmers were interviewed. Most interviews (N = 16) were conducted in spring 2024, with two additional interviews in October 2024. Interviews lasted between 2 and 5 h and were held on-farm. All interviews systematically included a visit of the grazing areas used by the farmers. This enables to refer to concrete situations and specific locations, and supported a more contextualised and grounded understanding of their decision-making processes.

A small subset of interviews (2 out of 18) was conducted by MSc students under the close supervision of the authors. All interviewers followed the same interview guide (Supplementary Material 1), and the interviews were recorded to ensure consistency and quality. All transcripts were analysed by the authors. The semi-structured format encouraged participants to describe their decision-making process in their own words, using practical references to objects, locations, or specific situations (Kauffman, 1996).

Each interview began with a brief overview of the farm’s structure and was then organised around five core topics: (1) farm characteristics, (2) role of rangelands in the livestock feeding system, (3) characterisation of pastoral resources, (4) decision-making processes related to grazing management, and (5) acquisition of local knowledge. Each topic was introduced by an open-ended lead question (identical across interviews), followed by optional supplementary questions to deepen and clarify responses (Table 1). The full interview guide is provided in Supplementary Material S1. The objective was to elicit farmers’ knowledge and to analyse how this knowledge is translated into decision-making processes and management practices.

The interviews were recorded and transcribed. Transcripts were then coded thematically using Taguette, an open-source qualitative analysis tool (Rampin and Rampin, 2021). Farmers’ narratives were organized into a matrix of themes and respondents, allowing systematic comparison of convergences and divergences across farms. The analysis unfolded in two stages.

First, we constructed a typology of farms. Based on a set of variables (Table 2), we grouped the farms according to the degree and manner in which rangelands were integrated into the feeding system. The classification was based on farm characteristics (e.g., utilised agricultural area, share of rangelands, flock composition, breeds) and management practices (e.g., stocking rate, duration of rangeland use, pastoralism rate, flock management, forage production, supplementation). It was constructed through iterative cross-farm comparisons, based on the visual clustering method of Bertin (1983), which is widely applied in qualitative research.

Second, we analysed farmers’ decision-making processes and underlying perception of the rangeland ecosystem within each type of farm. This second step focused on how farmers interpret their environment and make their decisions about grazing management on rangelands. In this study, “objects” refer to the elements of the pastoral system that farmers actively monitor in their daily practice, and that serve as reference points to interpret their environment and guide their decision-making (e.g., flock condition, resource availability, vegetation characteristics). Farmers’ discourse was analysed to identify: (i) the objects monitored, (ii) the indicators used to describe each object, and (iii) the strategies implemented to manage the complexity and variability of the pastoral ecosystem. This approach allowed us to explicitly link the objects monitored by farmers, the indicators they use to describe them, and the resulting decision-making processes.

The 18 farms in this study displayed a wide range of structural characteristics and grazing management practices, consistent with our objective of capturing diversity in pastoral systems. The details for each farm are provided in Supplementary Material S2. In the following sections, farm codes are used to link quotes to specific farms and groups. Farm codes start with “A” for Ardèche, and with “H” for Hérault.

All farms included shrubby or wooded rangelands in their grazing area. Their relative importance varied from around 50% of the utilised agricultural area in the most input-dependent farms to nearly 100% in the most extensive ones. Flock sizes varied from fewer than 50 goats to over 400 sheep, and stocking rates changed accordingly. Grazing management practices varied across farms. Shepherding was practised on most farms (12/18), either year-round or only in the grazing season, while four farms relied mainly on fixed or mobile paddocks. Patterns of mobility ranged from daily grazing circuits to seasonal transhumance across altitudinal gradients. Production objectives also differed across farms. Eight farms were specialised in dairy products (mainly goats), while ten focused on meat production with hardy local sheep breeds. Dairy farms generally displayed higher levels of intensification of the feeding system (more conserved forage and concentrate), whereas extensive meat farms relied more on local rangelands. This diversity in farm structure and management provided a robust basis for the typology presented below.

The typology revealed three types of farms (Table 3): P− (opportunistic rangeland users), P (adaptive users), and P+ (strategic users). These types differ not only in their level of reliance on rangelands, but also in how farmers interpret and manage resource variability.

P+ farms (N = 3) were characterised by a high pastoralism rate (∼90%), year-round rangeland use, very low supplementation, and 100% of their utilised agricultural area (UAA) consisted of rangelands. They relied on shepherded grazing, raised flocks composed of hardy local breeds (e.g., Caussenarde des Garrigues sheep, Rove goats) and accepted changes in animal performance, which had a limited economic impact due to low feeding costs. As one farmer [H7] noted: “I’m not looking for prolificacy. If the resources are there, great. If not, we don’t push it. They [the goats] self-regulate their intake.” P+ farmers displayed risk tolerance and valued self-sufficiency most.

P farms (N = 11) were characterised by moderate pastoralism rates (∼70%), intermediate rangeland use (∼10 months), mixed grazing strategies (paddocks and shepherding), and intermediate supplementation levels. The share of rangelands in the UAA averaged around 85%. P farmers relied on rangelands for most of the year, but practiced targeted supplementation (e.g., hay, alfalfa) during critical phases such as lactation. Some also (8/11) practiced seasonal transhumance to avoid summer shortages. This strategy accommodated both hardy meat sheep and more productive dairy goat systems. As one interviewee [H1] explained: “Most of the time, I use grazing resources as much as possible—especially before critical periods like lambing. When forage is lacking, I compensate with supplementary feed.” P farmers favoured grazing but adjusted the feeding system to secure their [moderate] production objectives against climatic uncertainty.

P- farms (N = 4) had a low pastoralism rate (<45%, often <40%), shorter rangeland use (∼6.5 months), moderate-to-high supplementation, and less reliance on long-range mobility. P- farmers typically developed paddocks or free-ranging systems, used more productive breeds and practiced on-farm production of conserved forage. Rangelands were grazed by animals with low nutritional requirements such as dry females. As one farmer [A4] put it: “I prefer to ensure a stable ingestion to prevent production drops, even if we could make better use of rangelands.” P- farmers followed input-dependent production models, prioritising animal performance and a tighter control of feed intake. These three groups of farms illustrate different strategies of rangeland use, which are associated with distinct decision-making processes analysed in the following sections.

Farmers expressed contrasting perceptions and objectives regarding the role of shrubby and wooded rangelands in their systems. These perceptions, which were unequally distributed among groups, ranged from highly positive valuations to more negative or constrained views.

First, all P+ and P farmers viewed rangelands as a pillar of the feeding system. They valued them for their low-cost contribution to forage self-sufficiency and to the resilience against climatic and economic uncertainties. As one farmer [A1] explained: “When grass is lacking, we can rely on woody plants. Animals eat the leaves, the fruits, and even if production or body condition declines a little, you can still maintain them.” Another farmer noted: “It’s really a matter of perspective. Once you understand that shrubs are a pastoral resource, you no longer want to remove, burn or shred them. You keep them—they provide shade and help maintain some grass during the summer.” The botanical and spatial heterogeneity of rangelands was perceived as a source of flexibility, enabling nutritional balance, stimulating intake and reducing parasite load. Another farmer [H7] noted: “It is the diversity of what they eat that matters. The more diversity you have, the better you make use of the resource.” A third farmer [H4] added: “These systems operate on a self-sufficient and cost-effective model, where external inputs are minimised.” For some farmers [A6, A7, H1, H4, H7], rangelands also contributed to the typicity and quality of animal products. These positive perceptions are associated with a greater reliance on rangelands and a more flexible approach to managing resource variability. As one farmer [H7] explained: “With climate variability, you’re never sure what resources you’ll have the following year, so you shouldn’t overuse them. And when there’s not enough, you have to go further to find it.”

In contrast, P− farmers considered rangelands as too variable and uncertain, requiring time-consuming constant observation and anticipatory adjustments. They highlighted the low or heterogeneous forage value of these areas, and the risks due to the presence of toxic plants. As a result, they deemed rangelands suitable only for animals with low nutritional requirements. As one dairy goat farmer [A4] explained: “I prefer to ensure a stable ration to avoid production drops. On rangelands I don’t think I can meet the needs of lactating females—at best they can be used for maintenance animals, but there is always the question of forage value, it’s too variable.” Another farmer [A5] stressed: “I think animals spend too much energy on rangelands—moving around, searching, dealing with climatic stresses. You cannot expect too much from rangelands, they are too uncertain.” P− farmers tended to respond to this uncertainty by increasing supplementation. As one farmer [A4] explained: “The more uncertain it is, the more I prefer to rely on supplementation to secure production.” Encroachment and difficult access further reinforced the perception of rangelands as unfavourable environments and, in some cases, as a burden within the farming system – areas that need to be grazed in order to avoid encroachment.

Finally, a number of farmers (12/18) attributed to rangelands a strong ecological and patrimonial value, beyond immediate feeding objectives. This perspective was not expressed by P− farmers and by two P farmers. For them, maintaining pastoral practices was important not only for flock feeding, but also for biodiversity conservation, landscape maintenance, and the preservation of local heritage. As one farmer [A1] stressed: “It is important to keep these lands open, not only for the animals, but for the landscape and heritage—this is also our role.” These farmers associated rangeland use not only with forage provision, but also with ecological and cultural services.

These contrasting perceptions of rangelands shape how farmers interpret resource variability and directly influence their decision-making processes regarding grazing management.

Our analysis of the decision-making process revealed that farmers mobilise four objects: landscape units, rangeland vegetation, pastoral resources and flock behaviour. For each of these objects, farmers referred to a set of indicators, which differed in relation to the time scale of grazing management (Figure 2). For instance, regarding the object “pastoral resources”, the daily grazing circuit could be managed on the basis of the spatial distribution of resource diversity, quality and abundance; the attribution of grazing units to animal groups for a period of time could be decided on the basis of the proportion of woody species and the seasonal availability of the pastoral resource; the setting of the whole feeding system could depend on the dynamics of the main forage resources in response to climatic conditions, past utilisation and their interactions.

The indicators used to describe each object and their use for decision making also differed between farm types. Illustrative examples of these differences, including verbatim quotes, are provided in Table 4.

Farmers differentiated landscape units on the basis of slope, altitude, microclimate, and accessibility; shade and water availability were considered as well, especially in summer. P+ farmers provided detailed and anticipatory descriptions of these features, adjusting grazing plans in advance according to expected changes in animal comfort and energy expenditure. P− farmers, in contrast, relied more on immediate, visual indicators—choosing units based on convenience or easily measurable signs (e.g., herbage height), with little integration of spatial or seasonal dynamics. P farmers occupied an intermediate position, with their strategies often shaped by the herd management mode: when animals were kept in fenced areas near the farm—especially around lambing—they made simple, short-term choices like P- farmers; when animals were herded, they relied on more anticipatory planning, though less refined than that of P+ farmers.

Generally, farmers described rangeland vegetation on the basis of its structure and composition, naming dominant plant species. They assessed vegetation dynamics in response to climate and grazing pressure, identifying key species (including woody plants) and indicators of rangeland health, such as plant cover diversity, regrowth potential, drought resistance, and presence of resilient species. Phenology and regrowth potential informed the ideal timing and intensity of use. P+ farmers showed fine-tuned knowledge of species composition, regrowth cycles, and palatability for livestock. P farmers used indicator species such as Brachypodium retusum (a low palatable grass) to decide on grazing rotations. In this study, rangeland “vegetation” refers to the biophysical characteristics of plant communities (e.g., structure and species composition), whereas “pastoral resources” refer to the fraction of vegetation that is accessible, selected, and consumed by animals, reflecting its functional role in feeding.

Pastoral resources were assessed with several criteria: nutrient content, palatability, toxicity risk, and forage diversity. P+ farmers stood out for their ability to accurately identify what constitutes a valuable resource for a given type of animal at a specific time of year, and to combine different plant species in ways that promote intake and nutritional balance. P farmers aimed to match high-quality grazing areas with key physiological stages of the animals. In contrast, P− farmers adopted a reactive strategy based on single indicators such as perceived forage quality, grass height, and immediate forage availability.

Animal behaviour provided real-time feedback on the quality and palatability of the forage. Farmers observed intake rate, movement patterns, and flock cohesion. Differences between animal species and breeds guided landscape use. Post-ingestive responses (e.g., rumen fill, milk production, body condition) were key to adjusting grazing strategies. P+ farmers showed a sophisticated interpretation of animal behaviour linking motivation, movement, and group cohesion. P farmers recognised the grazing preferences of different breeds—such as a tendency to favour certain plants or avoid steep terrain—and adjusted grazing timing and location accordingly. P− farmers used animal performance (milk yield, body condition) as delayed feedback, often reacting after issues had emerged.

The four aforementioned objects played complementary roles in decision-making, for all farms. The key differences between farms lay in how information about these objects was combined and interpreted. Table 4 illustrate a gradient of interpretive strategies, from reactive and fragmented to integrated and anticipatory approaches to grazing management. P− farmers tended to rely on simple, isolated cues (e.g., herbage height, milk yield), while P+ farmers integrated multiple indicators across dimensions to construct anticipatory strategies. P farmers occupied an intermediate position, shifting between simplified and integrated decision-making processes depending on the production cycle, resource availability, and labour constraints. Accordingly, Table 4 highlights distinct ways of engaging with uncertainty, autonomy and ecological variability through the integration of information across objects. This gradient—from fragmented to integrated interpretation—shows that attention should not only be devoted to which indicators are used, but also to how they are assembled into structured decision-making processes.

The objects observed and monitored by the farmers interviewed in our study — landscape units, rangeland vegetation, pastoral resources and animal behaviour — are consistent with observations from pastoral systems worldwide (Mongolia: Fernandez-Gimenez, 2000; Ethiopia: Oba and Kotile, 2001; Spain: Oteros-Rozas et al., 2013), and reflect the type of local ecological knowledge (LEK) long associated with adaptive rangeland management (Berkes et al., 2000). In particular, knowledge related to vegetation and pastoral resources typically includes the ability to identify plant species, understand their spatial distribution and seasonal dynamics, and assess their palatability for different types of livestock and at different times of the year. Such knowledge illustrates how pastoralists do not only observe vegetation as a static entity, but interpret it through its functional role in animal feeding.

Another common aspect is the use of animal behaviour cues in order to assess the amount and quality of the pastoral resources available, especially by farmers who practice shepherding. As Molnár (2017) summarised, “[pastoralists] see the grass through the mouths of [their] animals.” Monitoring and interpreting animal behaviour is an efficient way of dealing with the diversity and complexity of rangeland vegetation, which rarely lends itself to standardized quantitative assessments of forage availability, such as sward height or biomass measurements. Animal-based cues were part of all the pastoral decision systems we documented, although the indicators used differed between farm groups and types of decisions. These ranged from simple measures, such as body condition or animal weight, to more fine-gained, long-term observations of animal’s dietary choices. In some cases, this enabled farmers to anticipate flock’s responses on rangelands and influence their movements and exploration of the environment.

The cross-analysis of studies carried out in different countries reveals that the choice of indicators and their utilisation for decision-making differs between and within groups of pastoralists. For instance, Pakistani and East African pastoralists integrate behavioural cues as well as information on seasonal forage availability, quality and climate variability to guide their mobility and grazing decisions (Jamil et al., 2022; Oba and Kotile, 2001). Likewise, Mediterranean, West African and Maasai pastoralists combine topography, vegetation phenology and animal responses—treating animals as ecological indicators of palatability and dietary needs—to manage grazing under labour and risk constraints (Mapinduzi et al., 2003; Meuret and Provenza, 2015; Oteros-Rozas et al., 2013; Tamou et al., 2018). In our study we identified three distinct decision-making systems among French pastoral farmers, based on different combinations of indicators and responding to different priority objectives (self-sufficiency, low-cost production, or production performance). Our results contribute to documenting this diversity by showing how indicators are combined and interpreted within structured decision-making processes, in line with recent efforts to better characterise pastoral decision systems (Reyes-García et al., 2020; Wilmer et al., 2018). Rather than providing a descriptive inventory of local ecological knowledge, our analysis focuses on how farmers mobilise, combine and interpret different types of information to support their decisions. This highlights that the key issue is not only which knowledge is available, but how it is operationalised within decision-making processes. By making explicit these links between indicators, interpretation and action, this approach offers a way to analyse decision-making processes and helps identify where local ecological knowledge and scientific information may complement and support the development of context-specific advisory approaches.

Our results highlight a strong adaptive component of pastoral decision systems, within each group of farmers. In fact, the indicators used to support decision differed depending on the time scale of grazing management decisions: daily, mid-term (seasonal) or long-term (multi-year). Reed and Dougill (2002) argue that the use of plural indicators is itself a strategy to engage with uncertainty. Between groups of farmers, we documented a gradient of adaptation—from short-term reactive adjustments in the least pastoral systems, to targeted anticipatory decisions, to proactive integrative strategies in the systems relying most on rangeland. Beyond this gradient, pastoral decision-making is a dynamic process in which farmers perceive signals from the pastoral ecosystem, interpret them through experience and knowledge, act, observe outcomes, and adjust practices accordingly.

Our study also identified shifts in decision systems, especially for farmers aiming at low-cost animal production with moderate use of rangelands: certain periods of the year and animal groups would be managed following a “performance first” logic, while others would be managed following a “low-input” logic. Thus, farmers—particularly those in the intermediate group—may shift between different decision-making logics depending on resource conditions, market pressures or flock needs. This complexity echoes recent work emphasizing that farming practices cannot be reduced to a single rationality, but rather reflect diverse logics, priorities and relationships to risk and resource variability (Blasi et al., 2023). Darnhofer (2010) analysed resilience in farming systems, and suggested that adaptability results not from fixed inherently resilient strategies but rather from the ability to shift between decision logics as contexts evolve. Pastoral farmers’ decision systems thus align with the principles of adaptive management and resilience thinking (Berkes et al., 2003; Folke et al., 2010). These results illustrate how adaptation is not only expressed through practices, but also through the ways farmers interpret and combine information in their decision-making processes.

The comparison of the three decision systems highlights how farmers engage with the complexity and variability of rangeland environments. In particular, we identified that the farmers who relied most on rangelands had a low aversion to risk and a high confidence in the ability of rangelands to adequately feed their flock; conversely, farmers who used rangelands only a few months per year, for dry females, aimed at animal production stability and considered rangeland vegetation as an unreliable source of low-quality forage. In another geographical and production context, Michaud et al. (2020) showed that some French dairy farmers distrusted grasslands, while others relied on them to strengthen autonomy.

The accessibility of information about the pastoral ecosystem appears to play an important role in the refinement of rangeland management and in the perception of pastoral resources as well. In our study, the farmers that held the most accurate and diverse knowledge and displayed the most sophisticated decision system were also those who practiced shepherding. Although the practice of shepherding does not guarantee a high level of pastoral knowledge (see: Gobindram et al. (2018) for Moroccan shepherds), it implies a constant presence alongside the herd, on the rangeland. Thus, it gives the opportunity to observe the pastoral ecosystem across days, seasons and years. Since pastoral knowledge builds largely on repeated observations and trial and error tests by the farmers themselves (Sharifian et al., 2023), shepherding represents an asset for a good understanding of the pastoral ecosystem and the development of refined management strategies. Conversely, farmers who spend less time on the rangeland and/or with their grazing herd (either because they do not have the time, or they are not interested) will struggle to put together a decent amount of information about the pastoral ecosystem, which might become a “blind spot” in the feeding system. Such limited knowledge supports a lack of confidence in rangelands. Although our study did not specifically address the origins of farmers’ knowledge, the interviews suggest that learning processes are embedded in life trajectories, combining experience, observation and social transmission complemented by technical advice, as reported by Fernandez-Gimenez et al. (2021).

Whereas previous studies have treated LEK mainly as a corpus of observations and heuristics (Abule et al., 2005; Dabasso et al., 2012; Fernandez-Gimenez, 2000; Oba, 2012), we interpret it here as a dynamic process in which indicators are interpreted and combined to guide action (Jouven et al., 2024). This perspective provides a useful lens to analyse pastoral decision-making and to better understand how knowledge is mobilised in practice. By moving beyond a descriptive approach, our study shows how farmers mobilise knowledge through indicators within their decision-making processes. It helps identify how and where scientific knowledge may complement farmers’ own decision-making in a context-specific manner, and highlights potential points of articulation between local ecological knowledge and scientific information.

The uneven distribution of farms across the three groups, with a majority of P farms, reflects both the diversity of systems encountered in the field and the nature of this intermediate group. Indeed, P farms combine characteristics of the two more contrasting strategies (P+ and P−), and may rely on different decision-making logics depending on the time of year, production stage, or resource availability. Although this group could have been further subdivided, the focus was on capturing contrasted decision-making processes, particularly by highlighting the more distinct strategies represented by the extreme groups.

In addition, the sample reflects key structural characteristics of the study regions, which are predominantly composed of small-scale small ruminant systems. Flock sizes are generally small (< 500 heads), and large-scale pastoral systems are not represented. This focus is also consistent with the research objective, as it allowed us to analyse decision-making processes of farmers directly involved in day-to-day grazing management. In larger or more complex systems, grazing management may be partially delegated (e.g., to hired shepherds), making it more difficult to access individual decision-making processes. The study also focuses on systems relying on shrubby and wooded rangelands, which are characteristic of the study areas and involve specific ecological dynamics and management constraints. This provides a relevant context to explore decision-making under heterogeneous and variable conditions. The results may also provide insights into simpler situations, although their transferability to other types of pastoral context remains to be discussed.

These results suggest that advisory approaches in pastoral systems cannot rely on standardised recommendations, but need to be tailored to the diversity of farmers’ decision-making systems. As shown by the contrasted strategies identified in this study, farmers differ in the indicators they mobilise, their interpretation of resource variability, and their relationship to rangeland variability. Supporting adaptation to climatic variability therefore involves not only providing technical solutions, but also strengthening farmers’ capacity to interpret and combine indicators within their own context. In this perspective, the articulation between scientific and local ecological knowledge should be context-specific and aligned with existing decision-making logics.