Pyrenean meadows in Natura 2000 network: grass production and plant biodiversity conservation

In semi-natural mountain meadows, yield and forage quality must be reconciled with plant biodiversity conservation. This study was performed to analyze the relationships between these three parameters. To quantify plant biodiversity and pastoral value (PV), phytosociological inventories were performed in 104 semi-natural meadows in the Central Spanish Pyrenees included in the Natura 2000 network. Forage yields were calculated and forage samples were analyzed for relative feed value (RFV). We identified two main types of meadows: (i) those that had “more intensive management,” relatively close to farm buildings, with little or no slope, dominated by grasses, with low plant biodiversity, high PV and yield, but low forage quality and (ii) those that had “less intensive management,” distant from farm buildings, on slopes, richer in “other forbs”, with high plant biodiversity and forage quality, but low PV and yield. Conservation policies should emphasize less intensive management practices to maintain plant diversity in the semi-natural meadows in the Pyrenees. The widespread view that “other forbs” have low nutritional value should be revised in future research. These species often are undervalued by the PV method, because their nutritional quality, digestibility and intake are poorly understood. Additional key words: phytosociological inventories; botanical composition; Shannon index; pastoral value; Ellenberg indices; relative feed value; “other forbs”. Corresponding autor: rreine@unizar.es Received: 13-06-13. Accepted: 05-02-14 Abbreviations used: ADF (acid detergent fiber); CP (crude protein); DDM (digestible dry matter); DM (dry matter); DMI (dry matter intake); ECA (ecological compensation areas); EU (European Union); HNV (high nature value); LU (livestock units); NDF (neutral detergent fiber); PCA (principal components analysis); PV (pastoral value); RFV (relative feed value); SC (specific contribution); SF (specific frequency); SI (specific index). Instituto Nacional de Investigación y Tecnología Agraria y Alimentaria (INIA) Spanish Journal of Agricultural Research 2014 12(1): 61-77 Available online at www.inia.es/sjar ISSN: 1695-971-X http://dx.doi.org/10.5424/sjar/2014121-4617 eISSN: 2171-9292 62 R. Reiné et al. / Span J Agric Res (2014) 12(1): 61-77 Meadows are reservoirs of plant and animal biodiversity and have agricultural and ecological benef its such as high persistence (Alard & Balent, 2007), high intake rates (Duncan et al., 2003; Baumont et al., 2008), positive effects on animal health, and high sensory and nutritional qualities of animal products (Farruggia et al., 2008). These seminatural areas are indicators of biodiversity in agricultural landscapes (Blackstock et al. 1999). Some countries have developed programs to conserve or restore the floristic biodiversity of meadows (Henle et al., 2008), which include f inancial incentives for farmers. To maintain or restore semi-natural meadows, mowing or grazing are integral aspects of the management to produce forage for livestock feed and to allow farmers to balance economic and environmental objectives; however, in Pyrenean meadows, the intensif ication of these practices can reduce plant biodiversity (Reiné et al., 2004). Henle et al. (2008) suggested that the abandonment of marginally productive but HNV farmland is a major source of biodiversity-related conflicts and Hodgson et al. (2005) found that high biodiversity is associated with areas where the livestock carrying capacity and marginal returns are low. The economic benefits of intensif ication are associated with a reduction in biodiversity and the acceleration of the ecological processes that drive species losses in grassland ecosystems (Hodgson et al., 2005). The quantity and quality of the forage of a meadow depends on the floristic composition, which is influenced by the environmental, topographical and geographical features (e.g., climate, soil, moisture, elevation, slope, distance to the main farm building), and the spatio-temporal aspects of plot management (e.g., mowed or grazed, fertilization, time of year) (Kirkham & Tallowin, 1995; Wellstein et al., 2007; Cop et al., 2009; Andueza et al., 2010). Thus, a combination of environmental, technical, socioeconomic and historical aspects determines forage production. The valuation of the forage yield of a meadow is simple when it is mown only once (for hay or silage), but is more complex when it is used by grazing livestock. As a measure of quality, the chemical analysis of the total forage mass is valid only if the field is mowed; it underestimates quality when the f ield is grazed because livestock usually do not eat plants or parts of plants that are of low quality (selective herbivory). Following Daget & Poissonet (1972), scientists began to assign each species an index of quality [specific index (SI)] that reflects its agronomic value (growth rate, productivity) and zootechnics (bromatological value, digestibility, attractiveness). The SI is advantageous because the toxic and unpalatable species that livestock do not consume when grazing are excluded from the calculation of pastoral value (PV); however, the SI used in the PV method does not consider differences that are associated with phenology (Ansquer et al., 2004). The objectives of this study were to analyze the relations between plant biodiversity, yield, and nutritive value of 104 semi-natural meadows, with different botanical composition and environmental conditions, all located within the Natura 2000 network (OJ, 1992) in the Central Spanish Pyrenees. Our hypothesis is that the most productive grasslands will also have a high nutritional value but lower plant diversity, so that in the future could decrease their environmental value. Material and methods

To maintain or restore semi-natural meadows, mowing or grazing are integral aspects of the management to produce forage for livestock feed and to allow farmers to balance economic and environmental objectives; however, in Pyrenean meadows, the intensif ication of these practices can reduce plant biodiversity (Reiné et al., 2004). Henle et al. (2008) suggested that the abandonment of marginally productive but HNV farmland is a major source of biodiversity-related conflicts and Hodgson et al. (2005) found that high biodiversity is associated with areas where the livestock carrying capacity and marginal returns are low. The economic benefits of intensif ication are associated with a reduction in biodiversity and the acceleration of the ecological processes that drive species losses in grassland ecosystems (Hodgson et al., 2005).
The quantity and quality of the forage of a meadow depends on the floristic composition, which is influenced by the environmental, topographical and geographical features (e.g., climate, soil, moisture, elevation, slope, distance to the main farm building), and the spatio-temporal aspects of plot management (e.g., mowed or grazed, fertilization, time of year) (Kirkham & Tallowin, 1995;Wellstein et al., 2007;Cop et al., 2009;Andueza et al., 2010). Thus, a combination of environmental, technical, socioeconomic and historical aspects determines forage production. The valuation of the forage yield of a meadow is simple when it is mown only once (for hay or silage), but is more complex when it is used by grazing livestock. As a measure of quality, the chemical analysis of the total forage mass is valid only if the field is mowed; it underestimates quality when the field is grazed because livestock usually do not eat plants or parts of plants that are of low quality (selective herbivory). Following Daget & Poissonet (1972), scientists began to assign each species an index of quality [specific index (SI)] that reflects its agronomic value (growth rate, productivity) and zootechnics (bromatological value, digestibility, attractiveness). The SI is advantageous because the toxic and unpalatable species that livestock do not consume when grazing are excluded from the calculation of pastoral value (PV); however, the SI used in the PV method does not consider differences that are associated with phenology (Ansquer et al., 2004).
The objectives of this study were to analyze the relations between plant biodiversity, yield, and nutritive value of 104 semi-natural meadows, with different botanical composition and environmental conditions, all located within the Natura 2000 network (OJ, 1992) in the Central Spanish Pyrenees. Our hypothesis is that the most productive grasslands will also have a high nutritional value but lower plant diversity, so that in the future could decrease their environmental value.
A total of 104 meadows were chosen based on floristic and management characteristics: provided that their vegetation was included in the Habitat Directive (OJ, 1992) and their management practices could be known. Following these two premises a stratif ied preferential sampling was applied in four valleys. The average size of the meadow was 6,187 m 2 (465 m 2 minimum and 33,237 m 2 maximum). At the sampling sites, elevation ranged from 953 to 1,657 m.a.s.l. and slope ranged from 0% to 70% (Table 1).

Management practices
Most of the meadows were cut once for hay in July and grazed twice a year (spring and autumn) by cows or sheep at stocking rates less than 1.4 LU ha -1 year -1 . Farmers used NPK inorganic, cattle slurry, or manure like fertilizers. Accessibility of the plots and distance to farm buildings influenced agricultural management, a gradient being noticed from intensively managed meadows (one cut, two grazing, and inorganic or slurry fertilizer) close to the village, to more extensively managed f ields (no cut, two grazing, no fertilizer) close to the forest (Reiné et al., 2004).

Vegetation measurements
The vegetation was sampled on 1-15 July 2008, immediately before the harvest, at the peak of aboveground production. Phytosociological inventories were recorded in a 100 m 2 plot in the centre of each meadow. We used the Braun-Blanquet (1965) species abundance-dominance scale (from + to 5). The average plant height was calculated for each plot. Species nomenclature followed Tutin et al. (1964Tutin et al. ( -1980. To estimate yield, one biomass sample was collected from within a 0.25 m 2 quadrat in each plot. Dry matter was calculated using samples that had been dried in a laboratory at 70°C for 48 h. Species richness was the number of species recorded in each inventory. The plant diversity of the grassland community was calculated using the Shannon information index, H' (Shannon & Weaver, 1949): H' = -Σ s i=1 p i ln p i , where p i is the proportion of individuals belonging to the i th species and s is the number of species in the plant community. Plant species were classified based on their functional traits, in: (i) as either grasses, legumes, or "other forbs"; (ii) as either therophytes, geophytes, hemicryptophytes, chamaephytes, or phanerophytes (following Raunkiaer, 1934); (iii) following Ellenberg et al. (1991), who devised a comprehensive system for describing the response of individual species of vascular plants to an array of ecological factors (e.g., light, moisture, pH, nitrogen). The functional traits of each species were based on the Plantatt database (Hill et al., 2004).

Pastoral value
The pastoral value (PV) was calculated for each plot using the Daget & Poissonet (1972) method. The fieldderived cover-abundance data were transformed into percentage cover values based on the Braun-Blanquet coefficients: + = 0.1%, 1 = 5%, 2 = 17.5%, 3 = 37.5%, 4 = 62.5, and 5 = 87.5% ( Van der Maarel, 1979). Those values are denominated specific frequency (SF) and, after adjustment, Σ SF i is typically >100. In all the inventories, vegetation cover was 100%. Specific frequency (SF i ) values were converted to the specific contribution (SC) of each species using the following equation: SC i = (SF i / Σ SF i )100. The specific index (SI) ranges from 0 to 5 and is based on the productivity, digestibility, and attractiveness of each grassland species (Daget & Poissonet, 1972;Gillet et al., 2002). To assign a SI to each of the plant species identified in the inventories, we used the database of the Department of Agricultural and Environmental Sciences, University of Zaragoza, Spain, and Roggero et al. (2002). PV was calculated using the following equation:

Chemical analysis
Forage samples were oven dried at 65°C for 2 days and ground in a mill (IKA MF10, IKA-Werke, Staufen, Denmark) to the point where the material could pass through a 1 mm screen. Crude protein (CP) concentrations (N ⋅ 6.25) were quantified using the combustion method in an elemental analyzer (Elementar Vario Max N/CN, Hanau, Germany). Concentrations of ash-free neutral detergent f iber (NDF) and acid detergent fiber (ADF) were quantified using an Ankom 200 fiber analyzer (Ankom Technol. Corp., Fairport, NY, USA). Relative feed value (RFV) is an index that combines important nutritional factors (potential intake and digestibility) into a single number, which provides a quick, effective method of evaluating feed value or quality. The RFV is calculated using the estimates of digestible dry matter (DDM %) and potential dry matter intake (DMI % of body weight) of the forage based on the ADF and the NDF fractions, respectively (Linn & Martin, 1999)

Data analysis
Data analyses were performed using SPSS-Statistics (IBM, 2010). Pearson correlation matrices were used to assess the relationships between variables. To limit the number of variables included in the correlation matrix, we used principal components analysis (PCA). We have performed two PCA: (i) between variables of plant families, plant biodiversity, Ellenberg indices and topography, and (ii) between variables of plant families, plant biodiversity plant height, yield, forage quality and PV.
To create a classification of the meadows based on floristic composition, the presence-absence data of plant species were subjected to a TWINSPAN analysis (Hill, 1979). Normality of the 16 variables (grasses, legumes, forbs, species richness, Shannon index, light, moisture, pH, nitrogen, mean plant height, yield, CP, NDF, ADF, RFV and PV) was tested using Kolmogorov-Smirnov test. Cover values of grasses, legumes and forbs, expressed as percentages were arcsine square root transformed following Van der Maarel (1979). Differences for these 16 variables among the TWINSPAN groups of plots were assessed using one way ANOVA. When significant differences were detected, a HSD post hoc Tukey Test was used to compare means (Zar, 1984). The dispersion of the nine TWINSPAN groups was evaluated using linear, exponential, and logarithmic regressions.
In the 104 plots, 182 plant species (32 families) were identified, of which 29 were grasses, 23 were legumes, and 130 were 'other forbs'. The average number of species (species richness) per plot was 33 and the highest was 51 species (Table 1). On average, 27.2% of the species were grasses, which covered 44.3% of the plots, and 55.4% were 'other forbs', which covered 37.2% of the plots. Percentage of species (17.4%) and coverage (18.6%) were lowest among legume species. The average Shannon index (H') was 2.55 (range = 1.66-3.24).
Three (G7, G8, and G9) of the "more intensive management" groups had high coverage of grasses and lower coverage by 'other forbs' (Fig. 4). The other "more intensive management" group (G6) was similar to the groups that were under "less intensive management." The less intensively and the more intensively managed meadows did not differ significantly in the coverage by legumes. Groups that had more intensive management had markedly lower species richness; however, the Shannon index did not differ significantly between the two types of management, although the values tended to be lower in the groups comprising meadows that had more intensive management. Most (76.5%) of the plant species were hemicryptophytes (Table 1), and 94.9% of the plots contained . Mean coverage of grasses, legumes, and 'other forbs', and the species richness and Shannon index of the nine groups (G) of meadow types in 104 plots in semi-natural meadows in the Spanish Pyrenees. G1 to G5: groups that had less intensive management. G6 to G9: groups that had more intensive management. ANOVA test. Columns with different letters differ significantly (p < 0.05, HSD Tukey test). Error bars correspond to standard deviation. hemicryptophytes, therophytes and chamaephytes only. The less intensively and the more intensively managed meadows did not differ signif icantly in life forms (sensu Raunkiaer, 1934). Meadows under more intensive management had significantly higher levels of moisture and nitrogen, and tended to have higher light indices and pH than did the meadows under less intensive management (Fig. 5).

Yield, forage quality, and pastoral value
Yield, forage quality, and PV were highly variable among meadows (Table 1). In general, the meadows under more intensive management had higher average vegetation height, yield, NDF, ADF, and PV, and lower CP and RFV than did the meadows under less intensive management (Fig. 6). A matrix of Pearson correlations (n = 20 variables) indicated that many of the variables were significantly correlated (Table 2). In the PCA, the first and second axes explained 44.27% and 13.42% (total = 57.69%) of the variance, respectively (Fig. 7a). In the second PCA, the first and second axes explained 43.58% and 16.90% (total=60.48%) of the variance, respectively (Fig. 7b). The most significant correlations obtained with data from the 104 plots (Table 2 and Fig. 7) are shown clustered into the nine groups, in Fig. 8. Acid detergent fiber (%) Figure 6. Mean plant height, yield, forage quality, and pastoral value of the nine groups (G) of meadow types in 104 plots in semi-natural meadows in the Spanish Pyrenees. G1 to G5: groups that had less intensive management. G6 to G9: groups that had more intensive management. ANOVA test. Columns with different letters differ significantly (p < 0.05, HSD Tukey test). Error bars correspond to standard deviation.

Discussion
identified nine species groups (Fig. 1), which were classif ied as one of two main types of plant communities (Fig. 2): (i) meadows under "less intensive management" (G1-G5), which were on steeper slopes and at higher elevations than were the (ii) meadows under "more intensive management" (G6-G9). In our study and that of Ansquer et al. (2004) in France, L. perenne, P. trivialis, Holcus lanatus and Anthoxanthum odoratum were associated with meadows under more intensive management and Agrostis capillaris, Cynosurus cristatus, F. rubra, and Phleum pratense were associated with meadows under less intensive management. In Auvergne, France, Andrieu et al. (2007) found that L. perenne, P. trivialis, H. lanatus and Poa pratensis were associated with the meadows that were under more intensive management,  and A. capillaris and F. rubra were associated with meadows under less intensive management.
In mesic semi-natural grasslands in SW Finland, A. capillaris, F. rubra, L. vulgare, P. pratense, Taraxacum sp. and Veronica arvensis were associated with grazed grassland (unmown and, therefore, under less intensive management) (Pykälä, 2005). In meadows in Switzerland, Salvia pratensis was an indicator species that differentiated less intensively managed meadows from more intensively managed meadows (Schwab et al., 2002).
In our study, the groups that comprised meadows that were under more intensive management had more coverage by grasses (Fig. 4) and higher Ellenberg indices of moisture and nitrogen (Fig. 5) than did the groups that comprised meadows that were under less intensive management. In the former, the meadows were on level ground or shallow slopes and, therefore, could accumulate water and be fertilized mechanically more easily than could the meadows on steeper slopes. Intensive management tends to promote a plant community in which tall grasses predominate, which limits the light available to shorter plants (Marini et al., 2008). In our study, coverage by grasses and Ellenberg light indices were significantly positively correlated (Table 2). Meadows in the G6-G9 groups were relatively close to the main farm building, had little or no slope, were fertilized using organic and inorganic fertilizers, had one harvest for hay or silage in summer, and were grazed by cattle in spring and autumn, when they move to or come back from the alpine grasslands (located between 1600 and 2500 m a.s.l.). Grassland management practices vary widely in their disturbance impact, from low-intensity pasturing without fertilizers to mowing several times a year for fodder production (Wellstein et al., 2007). In France, Andrieu et al. (2007) found that the soils of the meadows that had been cut for hay or silage were more N-rich than were the meadows that were grazed, only, because the cut meadows received more fertilizer. In our study, the meadows in groups G1-G5 tended to be distant from the main farm building, on slopes, had low moisture, less frequently fertilized, and were exploited more for grazing than for mowing. As in our study, in seminatural grasslands in SW Finland that were grazed only, grazing increased the number of species associated with nitrogen-poor soils and low soil moisture (Pykälä, 2005).

Plant biodiversity
In the 104 meadows of our study, the average number of plant species was 33 (Table 1). In mountain meadows in France, average specific richness ranged between 25 and 38 (Gibon et al., 2004) and, in mountain pastures in Switzerland, the average was 30 (Meisser & Troxler, 2007). The mean SI of the meadows in the Pyrenees (2.55) was substantially lower than those observed in Switzerland (mean = 3.34), what can show a less balanced vegetation in our study.
Plant biodiversity (species richness and Shannon index) was highest in the meadows that had the least intensification (Fig. 4). Marini et al. (2008) suggested that, to preserve plant diversity in alpine meadows, conservation policies should promote extensive management practices. In hay meadows in northeastern Switzerland, land management had a strong influence on species richness, and the sites under extensive management had the highest species richness and diversity (Schwab et al., 2002). In our study, the meadows that were under less intensive management were more used for grazing than for mowing. Grazing increases species richness by increasing the diversity of microhabitats through selective consumption, trampling, urination, defecation, and by livestock acting as dispersal agent (During & Willems, 1984;Wellstein et al., 2007).
In the meadows of our study, high plant biodiversity was associated with plots on slopes and at the highest elevations (Fig 6a). The lower fertilization and the presence of microclimates in these topographical conditions interact to explain the biodiversity gain. In Finland, steep slopes that have poor soils experience extreme microclimatological conditions, which can increase species diversity by enabling less competitive plant species to persist . Marini et al. (2008) found that site features such as steep slopes were correlated with plant species richness in alpine meadows, and the effect was greatest in the meadows that had been mowed infrequently.
In the Pyrenean meadows, plant biodiversity was negatively correlated with the amount of grasses cover and each of the Ellenberg indices, and positively correlated with the amount of 'other forbs' (Fig. 7a and Table 2). Intensive land management, which entails high fertilization and heavy exploitation (mowing and grazing), favours the most competitive species; namely, those that have high growth rates and large size; e.g., grasses. Moderately intense exploitation allows more of the less competitive and, typically, slower growing species to persist. The Ellenberg indices of nitrogen and pH indicated that plant biodiversity was lowest on fertilized soils (meadows under intensive management). In alpine areas in north-eastern Italy, species richness tended to be highest in the meadows that had low management inputs and the least fertilization, which reflected the detrimental effect of intensive management on plant diversity (Marini et al., 2008). Increases in the use of fertilisers have been significant factors in the decline in grassland phytodiversity (Wellstein et al., 2007). Fertilizers increase biomass production, but reduce diversity (Di Tommaso & Aarssen, 1989;Fischer & Wipf, 2002) because they favour grasses and legumes at the expense of characteristic species; e.g., forbs (Fischer & Wipf, 2002). In our study, the amount of 'other forbs' and the Shannon index were positively correlated (Fig. 8a).
Intermediate levels of disturbance favour high species diversity (Grime, 1979;During & Willems, 1984;Wellstein et al., 2007). High financial returns, the objective of agricultural production, are correlated with a decline in biodiversity at intermediate-to-high productivity; which is consistent with the 'hump-back' model of Grime (Hodgson et al., 2005). In our study, we could consider that the meadows with 'less intensive management' are under these intermediate levels of disturbance.
The latest common agricultural policy (CAP) of the EU includes high nature value farming systems, and member countries will have to submit an agricultural systems catalogue that includes a reasoned explanation of the natural values of agricultural regions and the factors that limit their sustainability. Probably plant biodiversity indicators will be used to certify farmland as high nature value. In semi-natural meadows, yield and forage quality must be balanced against the need to conserve biodiversity, even if farmers perceive this as an inconvenience, initially. The policy should insure that farmers recognize the many benef its of plant biodiversity including an increase in the persistence of meadows, an increase in food intake by grazing animals because they can forage selectively, nutritional advantages, the benefits to animal health and the sensory qualities of animal products (e.g., milk, cheese, meat). In Spain, financial incentives should be used to encourage farmers to conserve biodiversity. In Switzerland, for example, each farmer is required to manage 7% of the farmland as ecological compensation areas (ECA), and the features that qualify as ECA include meadows, traditional orchards and hedgerows (Henle et al., 2008).

Yield and nutritive quality of forage
The forage yields of the meadows assessed in our study varied widely (1919-8880 kg DM ha -1 ), which paralleled the variability in plant height (0.3-1.6 m) ( Table 1). The meadows that were under 'more intensive management' (G6-G9) had the tallest vegetation and the highest yields (Fig. 6). In addition, yields were positively correlated with the amount of coverage of grasses and negatively correlated with plant biodiversity (species richness and Shannon index) and with the amount of coverage by 'other forbs' (Fig. 7b). Fertilizers increase yield because grasses are favoured at the expense of many 'other forbs'.
Forage quality varied widely (Table 1). CP ranged from 6.87 to 17.09 and RFV ranged from 78.58 (quality standard = 4) to 153.27 (prime quality). CP and RFV were correlated with the amount of coverage by 'other forbs' (Fig. 7b, see also Fig. 8b) and with plant biodiversity (species richness and Shannon index) (see also Table 2). Thus, the meadows that had the least intensive management (G1-G5) had the highest quality forage (Fig. 6). Furthermore, forage quality (CP and RFV) was negatively correlated with the amount of coverage of grasses and with yield; i.e., the forage of the meadows that were under intensive management had less quality (see Fig. 7b, and also Fig. 6), and was negatively correlated with forage quantity (Fig. 8c).
RFV is a function of digestible dry matter (DDM) and dry matter intake (DMI), which are inverse functions of acid detergent fiber (ADF) and neutral detergent fiber (NDF), respectively. Thus, high NDF and ADF are negatively correlated with RFV, which was apparent in the most intensively managed meadows in the Pyrenees, where grasses predominated (Fig. 6, Table 2). At similar stages of growth, typically, grasses contain higher concentrations of NDF and ADF than do legumes and 'other forbs' (Vázquez de Aldana et al., 2009). Concentrations of fiber are higher in the leaves and stems of grasses than they are in legumes (Buxton & Redfearn, 1997). In legume leaves, few tissues develop thick secondary walls; consequently, the total cell wall concentration of legume leaves does not increase with maturity as dramatically as it does in grass leaves (Jung & Engels, 2002). In grasses, the development of the stem at flowering involves a rapid accumulation of cell wall material; however, in legumes, material is added to leaves and stems, concurrently (Nordkvist & Aman, 1986). Typically, an increase in dietary NDF has a negative effect on the amount of DM consumed by cows (Allen, 2000).
The digestibility of 'other forbs' can be high (Table 2 and Fig. 8b) and varies relatively little over time (Bruinenberg et al., 2002, Andueza et al., 2010; consequently, the nutritional value of biologically diverse meadows remains stable throughout the growing season . Biodiversity allows grazing livestock to forage selectively, which favours high intake rates (Bruinenberg et al., 2002;Duncan et al., 2003;Baumont et al., 2008).

Pastoral value (PV). The case of the 'other forbs'
The average PV of the meadows in the Spanish Pyrenees was 50, which was 20% higher than the average in mountain meadows in Switzerland (Gillet et al., 2002;Gillet, 2008), although PV varied widely (23.65-74.48) ( Table 1). In other mountain pastures in Switzerland, PV ranged between 38 and 52 (Meisser & Troxler, 2007). In our study, PV was positively correlated with the grasses, plant height and yield, and was highest in the meadows that were under more intensive management (Fig. 7b, see also Fig. 6). PV can be a bromatological index of pasture quality (Daget & Poissonet, 1972); however, in our study, PV was negatively correlated with CP and RFV (Table 2, Fig. 7b, Fig. 8d). Furthermore, PV was negatively correlated with plant biodiversity and with 'other forbs' (Table 2, Fig. 7b and Fig. 8e). In wet grasslands in West Africa, PV and biodiversity were negatively correlated (Botoni-Liehoun et al., 2006). Grazing livestock can choose from dozens of plant species, but their preferences are not correlated with the SI (Agreil et al., 2004). In our study, PV was positively correlated with yield, but negatively correlated with forage quality.
In our view, the criteria used in the PV method to assign each species an index of quality (SI) are 'agronomic' and many of the species traditionally considered 'weeds' (typically, designated 'other forbs') are often underestimated or assigned an SI = 0 because their nutritional qualities, digestibility, and intake are poorly understood . Farruggia et al. (2012) used metabarcoding (sequencing plant DNA fragments) to identify in the faeces of grazing cattle many plant taxa that had an SI = 0. Some of those taxa can contribute to the stability of the feed value of multispecif ic meadows (Alard & Balent, 2007), provide the nutritional benef its of secondary compounds such as condensed tannins (Ramirez-Restrepo & Barry, 2005), which reduce the solubilization and degradation of proteins by rumen microorganisms (Min et al., 2000), phenolic compounds (including tannins) that provide insecticidal and antiparasitic benefits, which confers resistance to diseases and encourages herbivory (Waterman & Mole, 1994), and chemical compounds such as terpenes and phenols, which provide positive sensory qualities to animal products such as meat and cheese (Farruggia et al., 2008). The widespread view that 'other forbs' have low nutritional value should be evaluated in future research.
In semi-natural meadows in the Spanish Pyrenees, some geographical and topographical features (elevation, slope, distance to the main farm building) have influenced environmental conditions (soil moisture) and management systems (mechanization, fertilization, mowing, and grazing). Collectively, those factors have influenced the floristic composition of the meadows and, consequently, the yield and quality of the forage. In meadows close to the main farm building, which were in areas that had little or no slope, experienced 'more intensive management' (easy mechanization, fertilization with organic and inorganic fertilizers, one harvest for hay or silage in summer, and livestock grazing in spring and autumn) and grasses predominated, plant biodiversity was low, PV and yield were high, but forage quality (CP and RFV) was low.
The meadows that were on slopes, at relatively high elevations, or distant from the main farm building were under 'less intensive management' (little or no fertilization, exploited more for grazing than for harvest) and had a high proportion of 'other forbs', plant biodiversity and forage quality were high, but PV and yield were low. Management options of the meadows, compatible with environmental conservation, are possible with small adjustments to traditional systems in many cases, rather than intensive management or abandonment of their use by livestock.