Introduction
Wild herbivores and human-introduced livestock are directly involved in the maintenance of the taxonomic, physiognomic and functional diversity of grasslands. They are involved in colonization processes via seed dispersal, flower and fruit consumption and the opening of gaps in the vegetation. They also influence local extinction processes, control the populations of dominant species and help to increase spatial and temporal heterogeneity (Milchunas et al., Reference Milchunas, Sala and Lauenroth1988; Huntly, Reference Huntly1991; Hulme, Reference Hulme1996; Peco et al., Reference Peco, Espigares and Levassor1998). Livestock management and long-distance dispersal mediated by domestic animals have been recognized as two of the most decisive factors in the history of the fragmented landscapes that predominate in a large part of Europe (Poschlod et al., Reference Poschlod, Kiefer, Tränkle, Fischer and Bonn1998; Higgins and Richardson, Reference Higgins and Richardson1999; Nathan, Reference Nathan2001; Römermann et al., Reference Römermann, Tackenberg and Poschlod2005).
Although epizoochory is accepted as a fundamental part of the formation of agrosilvopastoral communities (Sorensen, Reference Sorensen1986; Poschlod et al., Reference Poschlod, Kiefer, Tränkle, Fischer and Bonn1998; Stiles, Reference Stiles and Fenner2000), few studies have quantified it with field experiments or lab simulations. Several studies of epizoochory in animals have been conducted under natural conditions, despite the methodological difficulties (Shmida and Ellner, Reference Shmida and Ellner1983; Liddle and Elgar, Reference Liddle and Elgar1984; Molinillo and Farji-Brener, Reference Molinillo and Farji-Brener1993; Fischer et al., Reference Fischer, Poschlod and Beinlich1996; Kiviniemi, Reference Kiviniemi1996; Olson et al., Reference Olson, Wallander and Kott1997; Mouissie et al., Reference Mouissie, Lengkeek and van Diggelen2005). Traba and Malo (Reference Traba and Malo2003) highlighted the difficulty of detecting and counting seeds on animal coats, which has possibly led to an underestimation of the dispersal potential of small-seeded species. Other authors have tried to minimize this problem by using species that facilitate seed handling, such as dogs (Heinken, Reference Heinken2000; Graae, Reference Graae2002) and mice (Kiviniemi and Telenius, Reference Kiviniemi and Telenius1998), although these mammals also affect the results. The same is true in the case of research involving the collection of seeds from game (Agnew and Flux, Reference Agnew and Flux1970; Heinken et al., Reference Heinken, Lees, Raudnitschka and Runge2001; Heinken and Raudnitschka, Reference Heinken and Raudnitschka2002; Heinken et al., Reference Heinken, Hanspach, Raudnitschka and Schauman2002), in this case due to post-mortem dragging. Another problem in these types of studies is that they do not reveal the retention time of the diaspores on the animal coats, an important aspect for knowing the plant's dispersal capacity. Recent experiments by Manzano and Malo (Reference Manzano and Malo2006) have found that seeds of four grassland species can travel long distances attached to the coats of transhumant sheep.
Many authors infer dispersal mode from the morphological traits of seeds and fruit. Epizoochory, in particular, has been associated with the presence of structures (hooks and awns) that enable seeds or fruits to become attached to animal coats. One classic example of this is Sorensen (Reference Sorensen1986). On the other hand, several field experiments have shown that species can be transported effectively on livestock hide without apparent adaptations to epizoochory, in some cases even with structures traditionally classified as anemochorous (Fischer et al., Reference Fischer, Poschlod and Beinlich1996; Heinken and Raudnitschka, Reference Heinken and Raudnitschka2002; Mouissie et al., Reference Mouissie, Lengkeek and van Diggelen2005).
Experiments have also been conducted to determine the retention ability of diaspores under standardized conditions (Traba et al., Reference Traba, Levassor and Peco2001; Couvreur et al., Reference Couvreur, Vandenberghe, Verheyen and Hermy2004; Tackenberg et al., Reference Tackenberg, Römermann, Thompson and Poschlod2006). The latter authors have studied the ability of propagules to adhere to the hide of different herbivores in more than 100 representative species of the north-west European flora, using a machine to ensure standardized shaking of animal coats. The experiment revealed a strong negative effect of diaspore mass. The expected positive effect of structures such as awns, bristles and hooks only appears in species with a seed mass >0.4 mg. Coat type also affects retention potential, with higher values in curly sheep wool in comparison to straight cattle hair, regardless of the coat position (vertical versus horizontal) (Tackenberg et al., Reference Tackenberg, Römermann, Thompson and Poschlod2006). Striving to apply these results to other floras and species, Römermann et al. (Reference Römermann, Tackenberg and Poschlod2005) proposed a regression model of attachment capacity, based on easily measurable indicator seed traits (diaspore mass and functional diaspore type), which explained 85% of the variability for sheep wool and 71% for cattle hair.
The present paper analyses the importance of diaspore morphology for animal coat attachment potential, using a sample of 14 abundant species from rangelands in central Spain that generally have smaller seed masses than the range of previously analysed species. We assessed: (1) the effect of diaspore mass, shape and dispersal structure presence on the attachment capacity of species representing a gradient of diaspore mass and shape; and (2) the effect of the animal coat type (sheep versus cattle) and position (vertical versus horizontal) on diaspore attachment capacity. Finally, we tried to validate the predictive model of Römermann et al. (Reference Römermann, Tackenberg and Poschlod2005) for this sample of small-seeded species.
Materials and methods
We measured the attachment potential of 14 abundant species from dehesa grasslands and scrublands in central Spain. The species selection covered a wide range of seed mass, shape and type of dispersal structures found in these grasslands (Table 1). Diaspores were collected in June and July 2002 from dehesa grasslands and scrublands in central Spain and stored dry at room temperature. Attachment measurements were conducted from October to December 2002. All measurements were conducted on the diaspore, i.e. on the dispersal unit including all attached structures. For convenience, all diaspores (seeds and fruits) are referred to as seeds in this paper.
Table 1 Mean values of seed weight, seed shape and presence of hard-to-remove dispersal structures in the species used in this study. Average percentages of attachment potential after 6 h in the coat-shaking machine are also included for different coats (sheep, cattle) and positions (vertical, horizontal)
Attachment potential measurements
Attachment potential (AtP) refers to the percentage of seeds still attached to a mechanically shaken animal coat after 0.5, 1, 2, 4, 8, 16, 32, 60, 120, 180, 240, 300, 360, 420 and 480 min. The latter two measurements were conducted after combing the coats to simulate animal behaviour when lying down or rolling on the ground.
The coat-shaking machine is described in Römermann et al. (Reference Römermann, Tackenberg, Poschlod and Knevel2004), and provides a realistic assessment of species retention under field conditions (Tackenberg et al., Reference Tackenberg, Römermann, Thompson and Poschlod2006). The experiments were carried out on sheep wool (Suffolk, wool length of approximately 3 cm, thick and curly) and cattle hair (Fleckvieh cattle, wool length of approximately 8 cm, fine, smooth and straight to slightly undulated). The coats were placed in two positions, horizontal and vertical. Horizontally fixed coats simulated the abdomen or back, and vertically fixed coats simulated the flanks or breast of the animal.
Before the experiment started, the coats were laid horizontally, and 100 seeds per species were spread evenly on them, as described in Tackenberg et al. (Reference Tackenberg, Römermann, Thompson and Poschlod2006). Four replicates of 100 seeds were used for the attachment potential measurements.
Seed traits
Seed mass and seed dimensions were taken from Sánchez et al. (Reference Sánchez, Azcárate, Arqueros and Peco2001). Seed shape was defined as the variance of the three main dimensions (first divided by length), following Thompson et al. (Reference Thompson, Band and Hodgson1993). Totally spherical seeds would have shape value of zero; shape values would increase with seed elongation. The species were also classified by presence/absence of appendages (awns, bristles, hooks, etc.).
Data analysis
A repeated-measures analysis of variance (ANOVA) was used to analyse the effect of coat type and position on attachment potential measurements. The results of this analysis were used to build different models to analyse the effect of seed morphology on attachment potential. Generalized linear models (GLM; McCullagh and Nelder, Reference McCullagh and Nelder1983) were used to analyse the effects of continuous (seed mass and shape) and categorical variables (presence/absence of appendages) on attachment potential after 6 h. The correlations among seed traits were analysed by Pearson correlations and the t-test.
Finally, the predictive models for attachment potential proposed by Römermann et al. (Reference Römermann, Tackenberg and Poschlod2005) were applied and compared with the values measured in this experiment using Pearson correlation coefficients. Attachment potential values were arcsine transformed, and seed mass was ln transformed to fulfil normality requirements. All statistical analyses were performed with SPSS 11.0 (SPSS Corp., Chicago, Illinois, USA).
Results
The repeated-measures ANOVA showed a significant effect of coat type and position, as did the interaction (Table 2). Attachment potential measured on cattle was low, compared to sheep wool for all types of seeds. Position also had a significant effect, but only in the case of cattle coats (F = 8.95, P < 0.006), with higher values of attachment potential in vertically set coats (Fig. 1). In all cases, temporal analysis of attachment potential showed that after the first few minutes, during which the non-adhered seeds fell off, the percentages of retention remained constant until the coat was combed (400 min), at which point further seeds fell off (Fig. 1).
Table 2 Effects of type of coat (cattle/sheep) and position (horizontal/vertical) on attachment potential after 0.5, 1, 2, 4, 8, 16, 32, 60, 120, 180, 240, 300, 360, 420 and 480 min on a coat-shaking machine for 14 plant species. F and P values of repeated-measures ANOVA
Figure 1 Average attachment potential measured after experimentally shaking sheep and cattle coats for different times (circles, sheep; triangles, cattle). The coats were placed in different positions in the shaking machine (vertical, white; horizontal, black). Bars indicate standard errors. Note that coats were combed at 400 min, before the last two measurements.
The results of the effect of coat type and position on attachment potential profiles (Fig. 1) led us to propose three independent models to assess the effect of diaspore shape on attachment potential: one for sheep coats, one for horizontal cattle coats and one for vertical cattle coats. Attachment potential decreased with seed mass for sheep coats and for vertical cattle coats (regression coefficient = − 5.214, P < 0.000, and regression coefficient = − 6.661, P < 0.05, respectively). The presence of dispersal structures had different effects on sheep and cattle coats. This parameter was insignificant for cattle coats and significant for sheep coats (F = 8.01, P < 0.009), where seeds with appendages showed greater attachment potential than seeds without appendages (93.3 versus 79.5%). Seed shape was only marginally significant in the case of horizontal cattle coats (regression coefficient = 124.6, P = 0.06), indicating a tendency for elongated seeds to have higher values of attachment potential than round seeds. No significant relationship was found between seed weight and shape (Pearson r = 0.15, P = 0.62), or between the presence of attachment structures and seed weight or shape (t-test = 0.09, P = 0.92 and t-test = 1.19, P = 0.26, respectively).
The tested model proposed by Römermann et al. (Reference Römermann, Tackenberg and Poschlod2005) yielded significant results in the case of the sheep coats for both the vertical and horizontal positions (Pearson r = 0.54, P = 0.04 and Pearson r = 0.59, P = 0.03, respectively), and was marginally significant in the case of the cow hide coat and vertical position (Pearson r = 0.48, P = 0.08).
Discussion
The effect of coat type and position
Attachment potentials measured on cattle were low in comparison with sheep wool for all types of diaspores. The detected differences suggest the influence of the vector species on the success of diaspore dispersal, as noted by Liddle and Elgar (Reference Liddle and Elgar1984), Kiviniemi (Reference Kiviniemi1996), Couvreur et al. (Reference Couvreur, Vandenberghe, Verheyen and Hermy2004), Mouissie et al. (Reference Mouissie, Lengkeek and van Diggelen2005) and Tackenberg et al. (Reference Tackenberg, Römermann, Thompson and Poschlod2006), in favour of thick, curly-haired coats, such as sheep. Römermann et al. (Reference Römermann, Tackenberg and Poschlod2005) found the same results as our study when comparing the percentage of seed retention on cow hide (hard, short hair) and sheep in an experiment using the same simulation device as the one used in the present study.
Some authors have considered the position factor experimentally to study accessibility and retention of diaspores on different parts of the animal body (Fischer et al., Reference Fischer, Poschlod and Beinlich1996; Kiviniemi, Reference Kiviniemi1996). From these studies, we know that the flanks are the areas where more seeds are retained under natural conditions. In our experiment, significant differences in attachment potentials between vertically and horizontally fixed coats were found only in cattle, where seeds were better retained on vertically than horizontally positioned coats. These results differ from those of Tackenberg et al. (Reference Tackenberg, Römermann, Thompson and Poschlod2006), who found no significant differences between positions. This discrepancy may be due to the differences in the data set and/or the type of analysis. Animal movements also affect diaspore retention, influencing the distribution of seeds along the routes followed by the livestock (Stiles, Reference Stiles and Fenner2000). In fact, our seed distribution curves show that after the initial moments of the experiment, the number of seeds on the coat remained practically constant until the combing treatment, leading us to deduce that animal behaviour is a decisive factor in diaspore depletion, as also mentioned by Liddle and Elgar (Reference Liddle and Elgar1984), Fischer et al. (Reference Fischer, Poschlod and Beinlich1996) and Stiles (Reference Stiles and Fenner2000).
The effect of diaspore traits
In the best-known studies of dispersal, one of the most widely used criteria for the classification of fruits on the basis of their adaptation to epizoochory has been the presence of structures that facilitate adherence, such as awns, bristles, hooks, etc. (Ridley, Reference Ridley1930; van der Pijl, Reference van der Pijl1982; Sorensen, Reference Sorensen1986; Hughes et al., Reference Hughes, Dunlop, French, Leishman, Rice, Rodgerson and Westoby1994). However, the results of the present study lead us to think that the presence of this type of structure is not strictly necessary for dispersal success, particularly in the case of small seeds. We found that the presence of appendages was only significant in sheep for which seeds with appendages do have high percentages of adherence (93.3%), although other diaspores without specific adaptations also adhere well (79.5%). Other morphological traits of the diaspores are involved in the dispersal process. Seed weight has the greatest effect on attachment potential, in this case negatively, as in the case of Römermann et al. (Reference Römermann, Tackenberg and Poschlod2005). This effect appears for sheep coats and vertically positioned cattle coats, but not for horizontally positioned cattle coats. In the latter case, only seed shape seems to have an effect on attachment potential, albeit marginally. In this case, long seeds seem to adhere better to the straight cattle hair, possibly because they penetrate the coat better. The fact that attachment structures have such little effect on attachment potential cannot be due to correlations with other variables used in the analysis. As seen above, there is no significant relationship between any of the analysed seed traits. Another noteworthy factor is the value of seed retention for certain shrubs that are generally not considered susceptible to long-term transportation via animal coats. The high attachment potential values of Lavandula stoechas in our experiment, for example, particularly in sheep coats (61.5 and 63.74% after 6 h in horizontal and vertical positions, respectively), suggest that epizoochorous dispersal may also be involved in the colonization of abandoned pastures in the case of gradual abandonment with sporadic use by livestock, as in many agrarian areas of the Mediterranean basin. These results complement those of Sánchez and Peco (Reference Sánchez and Peco2002), which illustrate the role of sheep, in this case via endozoochorous dispersal, in the colonization of new habitats by Lavandula stoechas thickets in Mediterranean grasslands of central Iberia.
Can we predict attachment potential capacity with easy-to-measure seed traits?
The model proposed by Römermann et al. (Reference Römermann, Tackenberg and Poschlod2005), for predicting the attachment potential of species of north-west Europe with easy-to-measure diaspore traits, has proven to be relatively robust, having been validated with a set of small-seeded species (range 0.01–8.7 mg) from the flora of Mediterranean grasslands and scrub. Despite the use of a dataset containing few species, the model was significant for sheep in both positions and partially significant for cattle in the vertical position. These results show the potential applicability of the Römermann et al. (Reference Römermann, Tackenberg and Poschlod2005) model, particularly for relative evaluations of attachment potential capacity on sheep flanks in floras other than those of north-western Europe. As the latter authors suggest, the model should only be used for large datasets to infer attachment potentials on ordinal scales.
The present paper also shows the importance of extending the range of traits in animals and diaspores that may be involved in epizoochory. This type of experimental approach helps to interpret long-distance dispersal and colonization phenomena, particularly in the fragmented and high-diversity habitats found across a large part of the Mediterranean basin landscape, where livestock movements are particularly important for their maintenance and restoration. The data from these experiments also facilitate retention ability comparisons between species from the same or different habitats, one of the basic aims of functional trait studies. However, as Nathan et al. (Reference Nathan, Perry, Cronin, Strand and Cain2003) point out, it must also be borne in mind that the epizoochorous dispersal process involves other factors apart from the attachment potential of the diaspores, such as seed production (Willson, Reference Willson1993; Bruun and Fritzbøger, Reference Bruun and Fritzbøger2002), height of release (Fischer et al., Reference Fischer, Poschlod and Beinlich1996; Graae, Reference Graae2002; Heinken and Raudnitschka, Reference Heinken and Raudnitschka2002), duration of dissemination (Fischer et al., Reference Fischer, Poschlod and Beinlich1996; Heinken and Raudnitschka, Reference Heinken and Raudnitschka2002) and plant abundance in the vegetation (Fox and Srivastava, Reference Fox and Srivastava2006). Mixed models that take into account all of these aspects are therefore needed to predict the epizoochorous dispersal potential of a plant species.
Acknowledgements
We thank Peter Poschlod for allowing us to use the shaking machine built by the technicians at Marburg University and Susanne Bonn for her technical assistance. The study was supported by the Spanish Ministry of Science and Technology (project REN 2003-01 562) and the same Ministry's FPI grant to Isabel de Pablo.
