Study Sites, Plants and Their Dispersers
This study was carried out during the fruiting seasons (August–December) in 2009 and 2010 in the Doñana National Park (510 km2; 37°9′ N, 6°26′ W; elevation 0–80 m), located on the west bank of the Guadalquivir River mouth, south-western Spain. The climate is Mediterranean subhumid, characterized by dry, hot summers (June–September) and mild, wet winters (November–March). Annual rainfall is very irregular, averaging 577 mm ± 39 SE, with 88·4% of rain falling between October and April (data from Natural Processes Monitoring Group, Doñana Biological Station, http://www-rbd.ebd.csic.es/Seguimiento/seguimiento.htm).
The Doñana area includes seasonally flooded marshes on a clay substrate (marshland) and pine Pinus pinea L. plantations and scrubs with other scattered trees on a sandy substrate (scrubland). We focused on the scrubland, which occurs in several patches varying in size and structure and which are isolated from each other by natural (e.g. marshes) and human (e.g. cultivations) barriers. Furthermore, Doñana has more than 2000 km of dirt tracks (62·5%) and firebreaks (35·5%), most of which were established c. 50 years ago.
To enable assessment of the whole fleshy-fruit shrub community of Doñana, we chose three study sites (called Reserva, Rocina and Matasgordas, respectively) separated by distances between 2·5 and 14 km. Reserva is covered by pine woods and a dense Mediterranean scrubland (covering 11·6 km2) dominated by Halimium halimifolium L., Rosmarinus officinalis L. and Stauracanthus spp. It has a relatively high presence of fleshy-fruit species such as Juniperus phoenicea subsp. turbinata (Guss) Nyman, Juniperus macrocarpa Sibth & Sm., Corema album L., Rubus ulmifolius Schott, Pistacia lentiscus L. and Phillyrea angustifolia L. (overall 0·14 ± 0·03 shrub m−2; mean ± SE). The scrubland area at Reserva has about 48 km of SLD. Rocina is a riparian woodland zone along a stream and surrounded by Mediterranean scrubland and croplands. Its scrubland area (3·5 km2) comprises scattered P. pinea with a dense understorey of Stauracanthus spp. Cytisus grandiflorus (Brot.) DC., and H. halimifolium. Here, fleshy-fruit plants such as Myrtus communis L., Asparagus spp, Arbutus unedo L., Vitis spp, R. ulmifolius, Chamaerops humilis L. and Olea europaea L. var. sylvestris are scarce (overall 0·07 ± 0·02 shrub m−2; mean ± SE). The local SLD system is 36 km long. Matasgordas is characterized by open Mediterranean scrubland (4·2 km2) dominated by scattered Quercus suber L., Fraxinus angustifolia Vahl and patches of H. halimifolium with a variable density, and a great amount of fleshy-fruit plants such as P. lentiscus, Pyrus bourgaeana Decne., C. humilis, P. angustifolia, R. ulmifolius, M. communis and O. europaea (overall 0·42 ± 0·08 shrub m−2; mean ± SE). This site contains 21 km of SLD.
In the Mediterranean basin, fleshy-fruit shrub species generally flower during later winter and spring (February–May) and produce drupes (e.g. P. lentiscus, R. ulmifolius, P. angustifolia) or berries (e.g. C. album, M. communis) that ripen between August and December (Jordano 1984a; Fedriani & Delibes 2009a). Depending on the species, each fruit contains generally from one to eight seeds, although R. ulmifolius frequently contain more than 20 seeds per fruit (Jordano 1995).
In Doñana, most of those plants are dispersed by mammals (Herrera 1989; Fedriani & Delibes 2009a,b), although some of them are also dispersed by birds (Jordano 1984b; Herrera 1995). Specifically, six frugivorous mammals are known to be local important seed dispersers: wild boar Sus scrofa L. (Matias et al. 2010), red deer Cervus elaphus L. (Perea et al. 2012), fallow deer Dama dama L. (Eycott et al. 2007), red fox (Fedriani & Delibes 2009a), Eurasian badger Meles meles L. (Fedriani & Delibes 2009b) and European rabbit (Delibes-Mateos et al. 2008). Genets Genetta genetta L. and Egyptian mongooses Herpestes ichneumon L. also occur in Doñana, but were not recorded in our surveys. Radiotracking studies (Fedriani, Palomares & Delibes 1999) and sign censuses (data from Natural Processes Monitoring Group, Doñana Biological Station) suggested that carnivores and rabbits tended to positively select SLD, whereas ungulates (boar and red/fallow deer) seem to avoid them. Furthermore, recent studies in the same area suggested that they also differ in the proportion of seeds they damage (Fedriani & Delibes 2009b; Perea et al. 2012). Therefore, and for the sake of simplicity, we classified these potential seed dispersers into three groups: ungulates (boar and red/fallow deer), carnivores (fox and badger) and rabbits.
Collection and Analysis of Faecal Samples
To assess the potential effect of SLD on different aspects of mammal-generated seed rains (i.e. faeces abundance and distribution, fruit consumption and seed damage), we surveyed four transects (500 × 2 m) for mammal faeces once a week, during both fruiting seasons in each study site (overall 12 transects, 6 km). At each site, two transects were established along SLD verges and two parallel to the SLD but at a distance of 60 m into the scrubland. The distance was selected to ensure the collection of ungulate faeces (a preliminary study suggested that ungulates avoided a buffer of around 30 m from SLD) and also that sampled transects fall within the same shrub community. Along each transect, we recorded the location and removed all faeces of target mammals. We assigned each mammal faecal sample to species on the basis on their shape, size and smell. For wild boar and carnivores, we assumed that all faecal samples were found. Deer and rabbit faecal pellets are scattered and therefore difficult to sample, we used the ‘pellet group’ as the sampling unit, defined as ≥30 pellets for deer and ≥50 pellets for rabbits, within a circular 50-cm diameter plot.
To attain a relative estimate of the number of seeds delivered in each habitat (SLD vs. adjacent scrubland) as well as the group-specific fruit consumption and seed damage, we analysed up to three faecal samples per disperser and survey (i.e. each transect sampled per week), depending on availability. Overall, we analysed 62·3% of collected faeces (n = 987). Faeces were dried and stored in paper bags. For their processing, they were soaked, carefully broken and cleaned. Then, we successfully identified and counted all seeds from fleshy-fruit plants, either damaged or intact. The number of damaged seeds was estimated by assessing the minimum number of pieces that made up a seed, considering the size of the whole seed and of each damaged piece and using a broad seed reference collection (Herrera 1989; Fedriani & Delibes 2009b; Perea et al. 2012). Although it is possible that mammals digested a fraction of ingested seeds, such fraction is likely to be small (Traveset 1998) and similar between habitats, hence this should not have a major effect on our results.
To examine whether SLD influence mammal-generated seed rains, we first examined for potential differences between habitats in the number of mammal faeces found per survey. To determine mammal fruit consumption and seed damage, we considered the proportion of analysed faeces containing fruit remains (i.e. seeds, pulp, skin or their fragments) and the proportion of damaged seeds regarding the overall number of seeds within the faeces, respectively. To assess the intensity of mammal-generated seed rain in both habitats, we considered the estimated number of unbroken seeds (mostly viable; Fedriani & Delibes 2009a) dispersed per survey as response variable. Finally, we also looked for potential differences between habitats in the richness and the diversity (estimated by the Shannon index) of dispersed plant species.
We evaluated potential differences among habitats and dispersers in our response variables by fitting generalized linear mixed models (by means of sas 9.2 GLIMMIX procedure; Littell et al. 2006). Negative binomial distribution and log-link function were assumed in all models, except for fruit consumption, which was fitted to a binomial distribution. For seed damage estimation, binomial distribution led to strong over-dispersion. Thus, we adjusted the model to a negative binomial distribution considering the number of damaged seeds per survey as response variable and introducing the total number of seeds found per survey as a random factor, to control for sample size variation.
In all mixed models, we considered the habitat (scrubland and SLD verges), the disperser group and their second-order interaction as fixed factors. When this interaction was significant, we performed tests for the effect of a factor at the different levels of the other factor (‘tests of simple main effects’) using the SLICE option in the LSMEANS statement (Littell et al. 2006). Year, the month of sampling (nested within year) and the transect (nested within site) were included as random factors to control for temporal and spatial heterogeneity. Adjusted means and standard errors were calculated using the LSMEANS statement, which estimate the marginal means over a balanced population (Littell et al. 2006). Whereas in a balanced sampling observed and adjusted means are usually similar, in unbalanced samplings (as it was the case of this study) observed and adjusted means may differ considerably.