Local Factors Rather than the Landscape Context Explain Species Richness and Functional Trait Diversity and Responses of Plant Assemblages of Mediterranean Cereal Field Margins

Arable field margins are valuable habitats providing a wide range of ecosystem services in rural landscapes. Agricultural intensification in recent decades has been a major cause of decline in plant diversity in these habitats. However, the concomitant effects on plant functional diversity are less documented, particularly in Mediterranean areas. In this paper, we analyzed the effect of margin width and surrounding landscape (cover and diversity of land use and field size), used as proxies for management intensity at local and landscape scales, on plant species richness, functional diversity and functional trait values in margins of winter cereal fields in southern Spain. Five functional traits were selected: life form, growth form, seed mass, seed dispersal mode and pollination type. RLQ and fourth-corner analyses were used to link functional traits and landscape variables. A total of 306 plant species were recorded. Species richness and functional diversity were positively related to margin width but showed no response to landscape variables. Functional trait values were affected neither by the local nor landscape variables. Our results suggest that increasing the margin width of conventionally managed cereal fields would enhance both taxonomic and functional diversity of margin plant assemblages, and thus the services they provide to the agro-ecosystem.


Introduction
Agricultural intensification is a worldwide phenomenon [1,2] adversely impacting biodiversity and ecosystem services within agricultural systems [3,4]. The landmark features of agricultural intensification include increased crop management intensity, e.g., high use of agrochemicals and short rotation schemes [1-3] and decreased landscape complexity resulting from a higher proportion of land designated for annual crops at the expense of land-use diversity, semi-natural habitats and field margins [3][4][5][6][7].
Field margins can contribute to alleviating the negative effects of agricultural intensification on biodiversity through their ability to sustain populations of pollinators, natural enemies and plant more frequent in complex landscapes, while medium-sized margins were more frequent in simple landscapes. Narrow margins were evenly located in both complex and simple landscapes (Figure 1).

Agricultural Intensification Effects on Species Richness, Functional Diversity and Functional Traits
Species richness, functional diversity for the full trait set and for seed dispersal mode were affected by margin width (Table 3, Figure 2). Wide (28 ± 10 species) and medium-sized margins (23 ± 9 species) had significantly higher species' richness than narrow margins (15 ± 6 species). Functional diversity, as measured for the full trait set, was significantly higher in wide margins than in narrow margins (Table 3, Figure 2), whereas for seed dispersal mode, it was significantly higher in wide margins than in medium and narrow margins (Table 3, Figure 2). Neither species richness nor any functional diversity variables were correlated with the landscape complexity gradient (PC1) or with any individual landscape variable ( Table 3).
The RLQ analysis did not reveal any shift in functional trait values in response to local and landscape variables. Random permutations of the rows of the R and Q tables, used to test respectively whether species presence is independent or not from the environment and from trait values, showed that both null hypotheses could not be rejected (p = 0.27 and p = 0.30, respectively). This joint result indicated a lack of support for an association between intensification variables and functional trait values. In concordance with the RLQ analysis, the fourth-corner analysis detected no significant association between individual trait values and margin width or any individual landscape variable (Table 4). Hemicryptophytes hemi 10 --- [33] Therophytes thero 47 --- [33] Growth form Forbs forb 44 --- [33] Grasses gras 14 --- [33] Pollination

Agricultural Intensification Effects on Species Richness, Functional Diversity and Functional Traits
Species richness, functional diversity for the full trait set and for seed dispersal mode were affected by margin width (Table 3, Figure 2). Wide (28±10 species) and medium-sized margins (23 ± 9 species) had significantly higher species' richness than narrow margins (15 ± 6 species). Functional diversity, as measured for the full trait set, was significantly higher in wide margins than in narrow margins (Table 3, Figure 2), whereas for seed dispersal mode, it was significantly higher in wide margins than in medium and narrow margins (Table 3, Figure 2). Neither species richness nor any functional diversity variables were correlated with the landscape complexity gradient (PC1) or with any individual landscape variable (Table 3).  The RLQ analysis did not reveal any shift in functional trait values in response to local and landscape variables. Random permutations of the rows of the R and Q tables, used to test  Table 3. Correlations of species' richness, functional diversity for the whole trait set (FDT) and for individual traits (FDS) of field margin plant assemblages with PCA first axis (PC1) and with landscape and local variables. Spearman's rank correlations (ρ) were used except for margin width class, for which correlations were determined by the Kruskal-Wallis (χ 2 ) test. In bold, p-values < 0.05.
For two quantitative variables, Pearson correlation coefficient (r) was used. For one quantitative and one qualitative variable pseudo-F was employed (F). For two qualitative variables, chi-square test was used (χ 2 ).

Discussion
In this study, we investigated the role of margin width and landscape complexity on species richness, functional diversity, and functional trait values of plant assemblages of margins of conventionally managed winter cereal fields in Southern Spain. Our results suggested that margin width had a significant effect on species richness and functional diversity of margin plant communities, whereas the effects of the landscape context appeared to be of minor importance. In addition, we showed that life form, growth form, seed mass, seed dispersal mode and pollination type in these plant communities were responsive to neither margin width nor landscape complexity.
According to our expectations, a positive effect of margin width was detected both on plant species richness and functional diversity. Unsurprisingly, wider margins were found to support higher taxonomic diversity than narrow margins. Quantitatively, margins wider than 1 m (medium and wide) harbored 40% more species than narrower ones [16][17][18]. Apart from a direct effect of width resulting from the increased area, higher diversity can also be gained through increased habitat heterogeneity in wider margins [37]. In parallel, a number of studies have provided evidence that narrow margins are strongly impacted by disturbances associated with the agricultural management of the adjacent field (e.g., herbicide drift and nutrient leaching) while wider margins gain a buffering capacity in front of disturbances [17,37,38]. Our study also showed that wide margins (>2 m) harbored a higher plant functional diversity than narrow ones (<1 m), i.e., allowed the coexistence of plants exhibiting dissimilar functional trait values, as found for the full trait space and for seed dispersal mode. These results suggest that taxonomic and functional diversity were not decoupled here, i.e., that the loss of species translated into losses in functions. This finding reinforces the view that Mediterranean field margins of conventionally managed cereal crops often shelter a low functional diversity, most likely due to their narrow character, which renders them disturbance-prone, and to a restricted species pool in the surrounding, mostly simplified landscape [30]. Indeed, the most frequent therophytes recorded here are pernicious weeds of winter cereal crops [39], which would suggest that narrower margins experience environmental conditions similar to those of crop fields.
The taxonomic and functional plant responses were mostly driven here by agricultural intensification at the field scale (margin width), rather than at the landscape level. Although landscape-scale effects on field margin plant diversity have been described in some studies [15,16,28,40], our results are in line with other studies suggesting an overriding role of local management on arable plant taxonomic and functional diversity [19,21,30,41]. This lack of landscape effects may have several plausible causes that could act in isolation or in combination: (i) as suggested earlier, an intensive management of cereal field margins that would override and mask the effect of the landscape context, (ii) a confounding effect of margin width and landscape complexity (i.e., wider margins were located in the more complex landscapes) that would hinder the detection of landscape-scale effects (Figure 2), and/or (iii) the general low representation of alternative habitats (Table 1) enlarging the plant species pool, so that local plant species pools may not have differed largely along our landscape gradient.
Finally, our expectation was that the width and landscape context of margins would act as "environmental filters" of functional traits in margin plant assemblages. Previous studies have provided evidence that agricultural intensification could select for arable plants both within the field [15,28] and in field margins [16,30]. However, in our study, the representation of individual functional trait values was unaffected by margin width or landscape features. There are two plausible explanations for this lack of response. First, as mentioned earlier, most field margins under focus here were strongly affected by disturbances, as suggested by the dominance of therophytes and short-lived perennial species (Table 2). This disturbance regime could have impeded the establishment of long-lived perennial plants, including woody species [42]. It is, therefore, not surprising that the response most commonly reported in the literature, i.e., an increase in perennial and woody species in wide margins and in a complex landscape [15,16,28] could not be detected in the present study. Second, most of the margins studied were located within simple landscapes, and the gradient of landscape complexity under study here may have been too short to filter out functional trait values.

Study Area
The study area was located in the Guadalquivir river basin (Andalusia, Southern Spain; Figure 3). Land use is dominated by cereal crops, followed by olive orchards and other annual crops such as sunflower or cotton [43]. Natural habitats, such as Mediterranean woodland and scrubland, represent a minor land use and are composed of small unconnected patches [22,43]. The climate in the study area is Mediterranean, with an average annual temperature of 18.6 • C and an average annual precipitation of 590 mm. Altitude ranges from 12 to 106 m a.s.l. Soils devoted to cereal cropping are alkaline with a texture varying from clayish to sandy loam.

Study Area
The study area was located in the Guadalquivir river basin (Andalusia, Southern Spain; Figure  3). Land use is dominated by cereal crops, followed by olive orchards and other annual crops such as sunflower or cotton [43]. Natural habitats, such as Mediterranean woodland and scrubland, represent a minor land use and are composed of small unconnected patches [22,43]. The climate in the study area is Mediterranean, with an average annual temperature of 18.6 °C and an average annual precipitation of 590 mm. Altitude ranges from 12 to 106 m a.s.l. Soils devoted to cereal cropping are alkaline with a texture varying from clayish to sandy loam.

Field Margin Selection
Ninety-four field margins adjacent to conventionally managed winter cereal fields were selected. Selected margins were located at least 2 km away from each other in order to avoid overlapping landscape buffers

Margin Width and Landscape Features as Measures of Agricultural Intensification
Margin width was considered as an indicator of the intensity of agricultural management at the local scale and was measured in-situ during the plant surveys. Margin width ranged from a few centimetres to more than three meters. Depending on width, the margins were categorized as narrow, medium-sized or wide margins (Table 5).
Six variables, used as proxies of the intensity of agricultural practices at the landscape level, were assessed within 1 km radius buffers around each margin [44] using the Geographic Information System SIGPAC (Sistema de Información Geográfica de Parcelas Agrícolas;

Field Margin Selection
Ninety-four field margins adjacent to conventionally managed winter cereal fields were selected. Selected margins were located at least 2 km away from each other in order to avoid overlapping landscape buffers.

Margin Width and Landscape Features as Measures of Agricultural Intensification
Margin width was considered as an indicator of the intensity of agricultural management at the local scale and was measured in-situ during the plant surveys. Margin width ranged from a few centimetres to more than three meters. Depending on width, the margins were categorized as narrow, medium-sized or wide margins (Table 5). Six variables, used as proxies of the intensity of agricultural practices at the landscape level, were assessed within 1 km radius buffers around each margin [44] using the Geographic Information System SIGPAC (Sistema de Información Geográfica de Parcelas Agrícolas; http://sigpac.mapa.es/ fega/visor/): (i) arable land cover, (ii) annual grassland cover, (iii) woodland cover, (iv) cover of human settlements (Table 5), (v) the Shannon-Wiener's diversity index for land use, and (vi) the size of the adjacent field. Cover of perennial crops (olive orchards and fruit tree crops) was strongly correlated with arable land cover (r = −0.87, p < 0.0001) and was thus not included as an additional explanatory variable.

Plant Surveys
Plant surveys were conducted during the ripening stage of winter cereals, between mid-April and early June, in 2009, 2010 and 2011. In each margin, plant species were recorded in an area of 20 m × margin width. The number of species recorded was used as a measure of species richness. Plant nomenclature followed [33].

Plant Functional Traits
A set of five functional traits related to plant persistence, resource acquisition and reproduction was selected: (i) Raunkiaer's life form: therophytes, geophytes and hemicryptophytes. This trait signals the strategy for plant persistence [42,45]; (ii) growth form: forbs and grasses. This trait is related to plant architecture, resource acquisition and tolerance to selective herbicides [44,46]; (iii) pollination type: entomogamy, anemogamy and autogamy; (iv) seed dispersal mode: zoochory, anemochory and barochory. Both pollination type and seed dispersal mode categories represent contrasting strategies for pollen and seed dispersal in space and time, and are related to colonization ability [32,47,48]; and (v) seed mass, a trait related to reproductive investment, seedling establishment ability and persistence in the soil seed bank [49,50].
Specific values for each trait were retrieved from existing databases (Table 1 and Appendix B). To avoid the influence of rare species, only species recorded in at least 10% of the margins were considered in subsequent analyses [51,52]. The average seed mass of cogeneric species was used for three species for which no seed mass data were available.

Functional Diversity
The Rao's quadratic entropy index [53] was used to measure plant functional diversity (FD) in field margins [54]. This index incorporates both the relative abundance of species and a measure of the pair-wise functional differences between species, by measuring species distance in the functional trait space: where s is the number of species, d ij is the distance in trait values between species i and j, and p i and p j are the relative abundances of species i and j. We used species presence/absence as our abundance measure, with present and absent species assigned an abundance value of 1 and 0, respectively. In this way, Rao's index is largely a measure of functional richness, i.e., the volume of niche space occupied by the species [55]. Rao's index was calculated for each single trait and for the full trait set (FDT).

Functional Diversity
A principal component analysis (PCA) [56], including the local and landscape variables (Table 1), was conducted after standardization (centering and scaling) in an attempt to produce synthetic variables representing the gradient of agricultural intensification across the 94 sampled margins. The association of each intensification variable to selected PCA axes was measured by the Spearman's rank-order correlation coefficient (quantitative variables) or the Kruskal-Wallis test (margin types according to width).
The influence of margin width type on species richness and functional diversity measures was tested by the Kruskal-Wallis test. The effect of each individual landscape variable and selected PCA axes on species richness and functional diversity was assessed by Spearman's rank correlation coefficient.
Two complementary three-table analyses, RLQ and fourth-corner, were further conducted to associate plant traits with local and landscape variables measuring the intensity of agricultural practices [57] (Appendix C). RLQ analysis is a multivariate technique that provides trait combinations that have the highest covariance with combinations of environmental variables [58]. Fourth-corner analysis tests relationships between individual functional traits and individual environmental variables [59]. Both analyses are thus complementary and require three tables, the environment table (R, the local and landscape variables measured on the 94 sampled margins), the species composition table (L, the species present in the sampled margins at ≥ 10% frequency) and the trait-species table (Q, the species values for the selected functional traits, Appendix B). The RLQ analysis performs a simultaneous ordination of the three tables in different steps. First, correspondence analysis (CA) and Hill and Smith analyses were used to analyze respectively the L, R (with row weights equal to the row weights of CA), and the Q (with row weights equal to the column weights of CA) tables. RLQ then calculated two separate co-inertias on the R-L and L-Q tables and identified axes in which the species scores were rearranged to maximize the covariance between the sampling units constrained by the explanatory variables (the R table), and the species scores constrained by the species traits (the Q table); this resulted in linear combinations of functional traits and the explanatory variables. A permutation model (model 6 with 999 permutations, as proposed by [60]) with Bonferroni correction for multiple comparisons was used to test the link between species traits and the environment. This permutation model encompasses two sub-models, models 2 and 4, which test the hypotheses that species presence is independent from their environment (row permutation of the R table) and from their traits (row permutation of the Q table), respectively. Both sub-models must be rejected to confirm the relationship between R and Q tables. The fourth-corner analysis then assessed the association between a pair of quantitative variables with the Pearson correlation coefficient, between a pair of qualitative variables with the Pearson chi-square and G statistic, and between one quantitative-one qualitative variable with the Pseudo-F and Pearson correlation coefficient. The significance of these relationships was tested by 999 permutations based on model 6 with Bonferroni correction for multiple testing.

Conclusions
Even though a total of 306 species were recorded in this study, it appears that the structure and management of the studied margins are currently the main factors limiting their plant taxonomic and functional diversity. Despite the selection of margins located along gradients of margin width and landscape complexity, plant assemblages of the selected margins were functionally not diversified, with a clear dominance of therophytes (of which many were pernicious weeds) and limited occurrence of perennial species. Such low functional diversity is most likely a result of the inability of these highly simplified habitats for buffering the effects of intensive field management practices. Promoting species richness and functional diversity of dryland cereal field margins in the study area could, therefore, be achieved by widening existing margins so that some woody species can establish and the proliferation of weed species can be limited. Further investigations are needed to establish the importance of margin width as a management tool aimed to conserve plant diversity in rain-fed cereal field margins.