Habitat modification and fragmentation through human activity are key drivers of change to plant–pollination interactions (Kuldna et al. 2009; Winfree et al. 2009). Many studies have focussed on the effects of these drivers on pollinator species richness and diversity, plant reproductive success, and community composition of both pollinators and plants (e.g. Eckert et al. 2010; Montero-Castaño & Vilà 2012; Stanley & Stout 2013). Recently, there has been a huge increase in the number of studies using food web approaches to quantifying impacts of habitat modification on plant–pollinator interaction networks (e.g. Forup et al. 2008; Vilà et al. 2009; Power & Stout 2011; Geslin et al. 2013). By analysing changes to the structure of interaction networks, specifically changes to parameters such as interaction symmetry and diversity, nestedness and connectance, authors have attempted to test how anthropogenic activity affects the stability of communities (Tylianakis et al. 2010). However, visitation networks do not necessarily reflect pollination events (King, Ballantyne & Willmer 2013), and few studies have linked community-level interaction networks with variation in plant mating (including outcrossing rates, paternity, identity of pollen donors), meaning that the relationships between network structure and plant reproductive function have not been clear (Gómez, Perfectti & Jordano 2011).
In this issue of Functional Ecology, Vanbergen et al. (2014) quantified the effect of livestock grazing in woodlands, and surrounding habitat context, on not only visitation networks, but also on plant mating systems. Many studies have shown that highly mobile pollinator species abundance and diversity can be influenced by the composition and configuration of the landscape around the site in which they are being sampled, although there are taxon-specific differences in responses (e.g. Rundlof, Bengtsson & Smith 2008; Kennedy et al. 2013). This can influence insect-pollinated plant communities (Power, Kelly & Stout 2012) and can be modified by local processes such as grazing intensity (Sjodin, Bengtsson & Ekbom 2008). However, few studies have analysed whether interaction network structure varies with landscape context (Kaartinen & Roslin 2011; Ferreira, Boscolo & Viana 2013), and none have taken the next step and examined the consequences for plant mating systems. Using the self-compatible marsh thistle, Cirsium palustre, as a model species, Vanbergen et al. (2014) were able to relate both local management (presence or absence of livestock grazing) and landscape context (habitat of the surrounding area) with network parameters, as well as outcrossing rates, paternity, inbreeding and relatedness among pollen donors of C. palustre.
In general, there is a trade-off associated with creating networks to test hypotheses about human impacts on plant–pollinator interactions: in order to make general conclusions and encompass site to site variation, a large number of sites are required; but sampling at each site is time-consuming, requires high taxonomic resolution, and the sampling method used can influence the structure of the network (Hegland et al. 2010; Gibson et al. 2011; Chacoff et al. 2012). Partners interacting in a network can also vary greatly from year to year (Petanidou et al. 2008). Thus, the temporal and spatial resolution of studies using a network approach is usually limited (Kaiser-Bunbury et al. 2010). In this case, the trade-off used by Vanbergen et al. (2014) was to sample eight sites for a relatively short period (10 mins), but each site was visited 20 times throughout a single season to quantify interaction networks. The authors demonstrated that this sampling was sufficient to represent structural properties (connectance), if not species richness. By combining network data with data on the mating system of C. palustre, this study revealed some interesting patterns.
For example, Vanbergen et al. (2014) showed that grazing increased species richness of floral resources (but not abundance of flowers) and resulted in more species interactions. If the study had stopped here, we might have concluded that there were positive impacts of grazing on pollination interactions and pollination services. However, visitation networks were less nested in grazed sites, and since nestedness confers robustness, grazed networks may be more vulnerable to species and habitat loss and further environmental changes (Memmott, Waser & Price 2004; Fortuna & Bascompte 2006). In addition, using microsatellites to genotype seeds of C. palustre to determine mating system parameters, Vanbergen et al. (2014) found that outcrossing rates in C. palustre decreased as networks became more connected in grazed habitats. This was despite the fact that there were more pollen donors in grazed habitats. It would be interesting to construct individual-based networks for C. palustre individuals in these habitats (cf. Gómez, Perfectti & Jordano 2011) to clarify mating patterns in different land-use contexts. Whether changes in outcrossing rates impacted on individual fitness in grazed habitats was not assessed, but is clearly another important question. Interestingly, there was no relationship between floral abundance, habitat (woodland) area or landscape structure and network parameters, and the authors attribute this to the fact that these networks were dominated by Diptera. This supports Kaartinen and Roslin's (2011) findings regarding food webs of herbivores and their natural enemies: the structure of food webs does not change with landscape context. Thus, although landscape context can affect individual species, and some interaction parameters, the overall structure of interaction networks appears stable. However, Fabian et al. (2013) found that forest cover and landscape heterogeneity affected host–enemy interaction network structure of trap-nesting Hymenoptera, and recent work has also concluded that landscape context can modify insect visitor network structure when networks are less dominated by Diptera (D.A. Stanley, D. Bourke & J.C. Stout, unpublished data). However, there is clearly more scope for studies to understand the relationship between landscape complexity and the structure of pollination interactions.
Characterizing interaction networks, particularly in species rich habitats, can be challenging, and network parameters can vary with sampling approach and network size (Gibson et al. 2011). Although correction for network size reduced, eliminated or reversed trends in network parameters, Vanbergen et al.'s (2014) study accounts for sampling effects and by combining interaction networks with plant mating systems to determine impacts of local and landscape scale processes, is a worthy contribution to the field. It clearly advances our understanding of the effects of anthropogenic disturbance on not just pollinators, but on the interactions they have with plants and the consequences for plant mating systems.