**1. Introduction**

Livestock represents a key element in society-nature interactions and is responsible for more than a third of global land use in a wide range of ecosystems [1]. Extensive, pasture-based ruminant and mixed crop-livestock systems provide 70% of milk and 60% of meat globally, utilizing 80% of all agricultural land [2]. In the Mediterranean, extensive, pasture-based ruminant systems have a long tradition dating back to antiquity. This form of livestock managemen<sup>t</sup> created characteristic landscapes, dominated by heterogeneous plant communities of forests, bushes, herbaceous undergrowth and grassland. A long co-evolutionary process generated "resilient ecosystems with a high species diversity, productivity and utility to society" [3]. As the specific environmental conditions in these regions limit intensive and specialized farming, extensive, pasture-based ruminant systems continue to shape many rural areas up until today [4]. In Europe, these types of ecosystems are considered High Nature Value (HNV) farmland (HNV farmland describes agriculturally used areas with high levels of biodiversity and is

one of 35 EU indicators for the integration of environmental aspects into the Common Agricultural Policy (CAP) [5]), with its highest proportions in Portugal, Spain and Greece [6]. In these regions, rough grazing biomass is transformed into high value products, traditionally by small, but increasingly also by large ruminants. The preservation of extensive, pasture-based small ruminant farming systems (SRFSs) is highly important for the protection of HNV farmland and rural communities throughout the Mediterranean [7]. In Greece, small ruminant farms produce 60% of milk and 65% of red meat for the national market, which is unique in Europe and indicates the sector's social and economic importance [8]. The present study addresses the sustainability of extensive SRFSs with a special focus on the Greek island of Samothrace.

Traditionally, SRFSs in Greece were, in most regions, characterized through transhumant activities. Since the beginning of the 20th century, nomadic lifestyles have been in decline, mainly morphing into semi-nomadic or sedentary extensive SRFSs. This model primarily uses common grazing land, provides supplementary feed and is characterized by low investment and productivity [9]. Today, sedentary extensive SRFSs shape rural Greece, and are currently threatened by a multitude of factors. While some regions su ffer from complete abandonment of grazing, others are heavily overgrazed [10]. Both trends lead to ecosystem degradation and demonstrate the current economic and social crisis this type of SRFS is experiencing. If grazing is abandoned, shrub and bush encroachment changes species composition and often leads to increased fire risk, while high grazing pressure mostly results in biodiversity loss and soil erosion [11]. Global agricultural industrialization has led to a decline in feed prices and production costs, even though transport distances are larger [1]. Consequently, prices for agricultural products have declined, while extensive, small-scale farms have an increasing di fficulty competing in the market. The demand for wool and skins has become so low that today many farmers prefer to dump them.

The EU Common Agricultural Policy (CAP) plays a special role in the transformation of grazing-based farming systems throughout the EU. In some regions, subsidy schemes supported the abandonment of grazing through the conversion of extensive pastures into forests or crop production; while in other regions, grazing was intensified through direct payments that initiated higher animal stocking rates [12]. Local socio-economic contexts and needs are often insu fficiently taken into consideration by EU-wide agricultural policies, resulting in mixed outcomes for farmers [13]. The aim of the present study is to highlight such a local socio-economic context on the Greek island of Samothrace, where the transformation of local agriculture was identified as the major driver for ecosystem degradation and widespread soil erosion [14–16].

The importance and multiple challenges faced by SRFSs in Greece and other Mediterranean regions, calls for a comprehensive research approach, focusing on environmental, social and economic aspects simultaneously [17]. In our case study, we aim to analyze the interdependencies of environmental, economic and social factors regarding the SRFS on Samothrace. We use the conceptual framework of socio-ecological systems research, as it builds a useful link between biophysical and socio-economic processes, by describing the exchange of materials and energy between society and nature [18,19]. The utilized mixed methods approach [20], integrates data on environmental, economic and social aspects of small ruminant farming from various sources and builds upon the long-term research project on the island. The integration of monetary flows expressed in relative prices complements the socio-metabolic approach, as it directly influences biophysical flows through farmers' behavior [21]. This approach allows us to derive sustainability indicators, assess socio-economic drivers, and define possible pathways for a sustainable future for agriculture on Samothrace. The study is guided by the following research questions: What factors contribute to and represent the current sustainability crisis of the SRFS on Samothrace? What are the socio-economic drivers for the regression of sedentary extensive SRFSs in Greece? What role does the EU Common Agricultural Policy (CAP) play, in the context of the current sustainability crisis of small ruminant farming on Samothrace? What could a sustainable future of small ruminant farming in the Mediterranean look like?

In Section 2, we outline the methodological approach by introducing the study site in Section 2.1, the conceptual framework in Section 2.2, the biophysical assessment in Section 2.3, the socio-economic assessment in Section 2.4 and the evaluation of uncertainty of input data in Section 2.5. Results are provided in Section 3. The Discussion in Section 4 contains chapters about the sustainability crisis of the SRFS on Samothrace in Section 4.1, the regression of sedentary extensive SRFSs in the Mediterranean in Section 4.2, the role of EU CAP in the changes affecting the SRFSs of Samothrace in Section 4.3, and the future of small ruminant farming in the Mediterranean in Section 4.4. Conclusions are provided in Section 5.

#### **2. Material and Methods**

#### *2.1. The Island Samothrace*

Samothrace stretches over 178 km<sup>2</sup> and is one of the very few hotspots of preserved archaic wilderness among the Greek islands. Its remote location in the north-eastern Aegean Sea, the pebbly nature of most beaches and often unclear land ownership, averted economic exploitation and mass tourism on the island. The 1611 m high mountain range Σα´*o*ς gives Samothrace its geomorphological character and shapes the distinct microclimates. While the northern side presents itself in lush green with old forest cover and numerous streams of drinkable water, the southern and western sides are shaped by a rather typical dry summer Mediterranean climate and vegetation. A large proportion of the island's terrestrial area is part of the Natura 2000 network, and since 2012, the island has been a UNESCO MAB ("UNESCO's Man and Biosphere (MAB) Program is an intergovernmental scientific program striving for the improvement of the relationship between people and their environment. The Biosphere Reserve concept started by a Task Force of UNESCO's Man and the Biosphere (MAB) Program in 1974, while the World Network of Biosphere Reserves (WNBR) was launched in 1976. Biosphere Reserves (BR) are terrestrial and/or marine areas that encompass valuable ecosystems and social communities that wish to combine the conservation of ecosystems with their sustainable use. They are established to promote and demonstrate a balanced relationship between humans and the biosphere" (www.sustainable-samothraki.net) candidate [14,22]. The island community of Samothrace is officially registered as 2840 people [23], but is subject to high fluctuations because many people leave the island in winter months or visit the island as tourists, seasonal workers or second homeowners. Of the 1000 economically active residents, 20% work as livestock herders and small-scale farmers. The secondary sector is relatively underrepresented at 10%, while the tertiary sector employs 60% and consists mainly of tourism services.

The development path of recent decades has led to a wide variety of environmental but also social problems which the island community currently must face. One of the major threats to local ecosystems was triggered by the transformation of the local agricultural system. Decades of overgrazing by sheep and goats resulted in biodiversity reduction and widespread soil erosion [15,16]. Since the mid-20th century, farms and farmers are declining, while the small ruminant population increased to unprecedented levels [24]. Increasing feed prices, dependence on subsidies, the lack of marketing opportunities and little cooperation among themselves, have caused local farmers to find themselves in an economic deadlock situation that now threatens the very existence of agriculture on the island.

#### *2.2. The Conceptual Framework*

For the present study, we defined the system of investigation as the small ruminant farming system (SRFS) and its most relevant socio-economic and ecological relations, shown in Figure 1. The green circle represents the natural, and the blue circle, the cultural sphere of causation, with the livestock and human population in the overlapping part. The SRFS is defined as the small ruminant population (sheep and goats), its metabolic requirements, its material output in terms of products, the small ruminant farmers and their monetary economy. Terrestrial ecosystems provide the net primary production (NPP) consumed by small ruminants. The SRFS exchanges goods and money with the

local population, including visitors. The political, legal and cultural framework is represented by rules and regulations of the Greek state, and the EU and local traditions. The EU provides agricultural subsidies through the Common Agricultural Policy (CAP), and the Greek state pays pensions to retired farmers. The local and visitor population receive money from external markets and through income from external sources (e.g., work or pensions). Wastes are not explicitly assessed in this study, but they are a relevant factor, especially regarding slaughtering residues and emissions.

**Figure 1.** Schematic representation of the studied system and important influential factors. Figure adapted from Petridis et al. and Fischer-Kowalski et al. [25,26].

The present study is part of a long-term research project on Samothrace, beginning in 2008. Thus, we were able to build upon previous research efforts and use data from various sources. The aim of covering environmental, social and economic factors resulted in the need to utilize a mixed methods approach [20]. As indicated by the grey boxes shown in Figure 2, we integrate quantitative and qualitative data from a survey with 23 small ruminant farmers, qualitative data from 12 expert interviews, public statistical data, and data from previous research on land cover dynamics [24,27]. For the assessment of biophysical flows, we utilize a bottom-up or stock-driven approach, to assess the biomass consumption of the SRFS on the island. Detailed documentation can be found in the Supplementary Data (SD) and Information (SI) file.

TheMaterial andMethods section is divided into the biophysical, in Section 2.3, and socio-economic, in Section 2.4, assessment of the SRFS system. The utilization of the metabolic small ruminant model in combination with various data sources requires a systematic assessment of uncertainty of the model and exogenous data [28]. For this reason, we conducted a sensitivity analysis in combination with a qualitative description, and when applicable, an uncertainty range of key input parameters in Section 2.5.

**Figure 2.** Mixed methods approach as applied to the SRFS (small ruminant farming system) and its socio-economic and natural environment.

#### *2.3. Biophysical Assessment of the Livestock System*

The assessment of biophysical conditions is based on the conceptual framework of socio-ecological systems research [18,19], and utilizes the methodological approaches of material and energy flow accounting (MEFA), and principles of human appropriation of net primary production (HANPP) on a local island level. The assessment was conducted in three ways:


The ODD (Overview, Design Concepts, Details) protocol is structured around *overview*, *design concepts* and *details* sections. Sections *overview* and *design concepts* are provided in Table 1. The *details* section is outlined below and includes the description of the *initial state* of the model and *model input* and *output data*.


**Table 1.** Overview and design concepts of the utilized modelling approach following the ODD (overview, design concepts and details) protocol.

The **initial state** of the utilized modelling approach is set by the official annual numbers of sheep and goats and their energy requirements, expressed in tons of dry matter and carbon content of biomass. The assessment of the feed intake of sheep and goats on Samothrace is based on GLEAM, which generally follows an LCA approach with the goal of assessing emissions of livestock production systems. GLEAM includes a method to derive feed inputs of small ruminants based on their energy demand, and production output through the utilization of the herd and feed ration modules (Supplementary Information: Figures S6 and S7). These modules were built in Excel by using the equations provided by the FAO in the GLEAM model description.

**Model input data** was derived from official statistics, survey data, literature and GLEAM model parameters. The herd module requires annual input data for the number of animals, live weights and ratios of cohorts, death, fertility and replacement rates, lambing/kidding intervals and litter size. The feed ration module requires input data for the daily milk production, annual production of fiber, feed rations and their average digestibility and gross energy content. A survey with 23 local livestock farmers was conducted in 2017–2018 to collect data on 176 parameters regarding flock characteristics, production, processing, grazing, feeding, land management, revenue and costs. Modelling of biophysical flows, live weights of animal cohorts, proportion of dairy animals, male to female ratio, lambing or kidding intervals, litter size and daily milk production were derived from survey data (for a detailed description of the selection process and sample see Supplementary Information Section 3). Death and replacement rates, average daily weight gains, average digestibility of feed ration and average gross energy content of feed ration were derived from region specific FAO data provided in the GLEAM model description. Total annual numbers of animals at the end of each year from 1993 to 2016 and their fertility rates were derived from official statistical data. The calculation of feed rations is based on the energy demand of animals in relation to the available amount of feed. Local feed production was derived from official statistical data on annual primary production [29]. Crop residues were calculated based on areas used agriculturally for crop cultivation and factors derived from literature [34]. Data from local traders was used to estimate annual external feed supply through imports. FAO data was used for the share of leaves in the diet. Total available feed was integrated to calculate the feeding ration of small ruminants in which the remaining feeding gap was assumed to be filled by grazing (Supplementary Data: Tables S3 and S4).

**Model output data** is generated for total feed demand for small ruminants in fresh grass, hay, crop residues, leaves and grains from 1993 to 2016, in tons of dry mass per year (tDM/year) and tons of carbon per year (tC/year). The model further calculates the share of feed demand that was covered

by imported feed, locally produced feed and grazed biomass from 1993 to 2016. By multiplying the number of milked animals by the duration of the milking season and the daily production potential, it was possible to estimate the potential annual milk production of all sheep and goats. Standard deviation values for average daily milk production (sheep: ±12%; goats: ±25%) were generated through integration of data from di fferent sources (Supplementary Information: Section 4). The herd module of GLEAM also allows for the calculation of the share of animals that is available for meat production in each cohort. A standard deviation of ±15% was defined. The modelled increase in production of milk and meat would cause a higher feed demand which is not considered in the results.

#### 2.3.2. Estimation of the Grazing Capacity of Local Ecosystems

For the assessment of the potential overutilization of grazing resources through sheep and goats from 1993 to 2016, we provide estimates for local net primary production (NPP), in combination with a trend derived from the assessment of the Normalized Di fference Vegetation Index (NDVI). Data on total available biomass for grazing was derived from Fetzel et al. [24], who utilized MODIS Net Primary Production (NPP) data for 9 CORINE land cover classes identified for Samothrace [35]. These land cover classes were grouped into 3 major land-cover types: "arable land", "natural forests", and "principally agricultural land with significant natural vegetation". For each land-cover type, the authors defined maximum biomass o ff-take levels to ensure that the feed supply estimates from local ecosystems are realistic and would not degrade essential resources. These net primary production (NPP) levels do not represent total aboveground biomass (NPPact) but refer only to the amount that can be grazed without continuous degradation of local ecosystems. For simplification, we refer to this level of net primary production as NPP in this study. The Normalized Di fference Vegetation Index (NDVI) trend applied to NPP values is based on previous research. The NDVI for Samothrace was calculated based on LANDSAT-datasets and their spatially discrete land-cover classifications in combination with a time-series analysis of continuous field data on biophysical ecosystem properties [27]. The NDVI trend for the years 1993 to 2016 was applied to the average NPP for the years 2000 to 2004 in order to derive annual NPP values for the covered period. As this approach is prone to relatively high uncertainties, we applied an NPP range of ±27% with regard to an uncertainty assessment for MODIS and NDVI data sources, derived from Jia et al. [36].

#### *2.4. Social and Monetary Assessment of the Local Small Ruminant Farming System*

Qualitative and quantitative surveys were conducted in order to integrate information about constraints and opportunities for agriculture on the island and data on the monetary economy of the small ruminant system. 12 qualitative expert interviews with 6 farmers (Expert 1–6), the dairy owner (Expert 7), 2 traders (Expert 8 and 9), 1 municipal employee (Expert 10), 1 local agricultural consultant (Expert 11) and 1 former vice mayor (Expert 12), were conducted between 2012 and 2018. A content analysis in regard to the past, present and future situation of small ruminant farming on the island was also conducted. Qualitative data from the survey with 23 small ruminant farmers was used to describe the sample (Supplementary Information: Section 3). Results of both qualitative analyses have been integrated into the discussion section. For the monetary assessment, retail prices for milk, cheese and meat, meat processing costs and feed costs for hay and grain, were asked for in the survey (Supplementary Information: Table S1 and Figure S1) and multiplied by the actual purchased or sold quantities of products to estimate annual expenses and revenue. Land, transport and farm utility costs were assessed in the survey per farmer, and average values applied to all small ruminant farmers on the island. For 2016, total revenues from milk, dairy products, meat, wool and subsidies are contrasted with total costs for feed, labor, land, transport, farm utility, processing, and the veterinarian. Total income or loss for the entire small ruminant production system and the average farmer was calculated.

#### *2.5. Validation and Uncertainty of the Utilized Model and Data*

The sensitivity analysis was applied to biophysical and monetary parameters. 23 input parameters were tested with a freely chosen ±10% factor and their effect on the 6 output variables, total feed demand [tC/year], sustainable grazed biomass [t/year], revenue [€/year], costs [€/year], income [€/year], potential milk production [kg/year], and potential meat production [kg/year] evaluated. Figure 3 shows the deviation of output variables with higher values than ±1.5%. The different colors represent the 6 different output variables in regard to the ±10% deviation of the input variables. The left side represents minus 10%, the right side plus 10% deviation of input variables. Transparent colors indicate a minus deviation, and full color bars represent a plus deviation of output variables. For a detailed description of the uncertainty evaluation of model input parameters see Supplementary Information: Section 6.

**Figure 3.** Results from a systematic sensitivity analysis, where each model input parameter was varied by +/−10%andtheeffectsonthemainindicator are thenplotted.
