Our results suggest that agroecosystems and associated marginal habitats contain both native plant species and potential environmental weeds; this highlights the balancing act that land managers face trying to ensure that management has positive effects on native diversity, while minimizing the risk of negative effects via promotion of weeds. Our main finding is that more native plant species, but fewer weed species, were observed in the ecotone between kiwifruit orchards and neighbouring native forest vegetation than were observed in the ecotones of co-occurring grassland agroecosystems (Table 1). Thus, for the spatial scale considered in our study (i.e. the 2 × 2 m plot level), there was no support for the view that kiwifruit orchards have greater weed forcing in ecotones (or forests) compared with nearby pastoral grassland systems (Fig. 1).
Changes in native and non-native species along the management gradient
Indigenous plant species comprised about a third of the species recorded in the study region, and the total abundance of these species increased by an order of magnitude from managed to ecotones or adjacent native forest (Table 2; see also Appendix S3). The plant community composition shifted in predictable ways from managed and boundary to ecotone and forest zones. For native species, this shift was from dominance by short-statured herbaceous species to dominance by ferns, shrubs and trees. Native species observed in managed and boundary zones were typically small-statured native herbs such as Hydrocotyle moschata, Euchiton involucratus and Cardamine hirsuta that persist in short-statured vegetation characterized by frequent disturbance. In ecotones, native species were dominated by a greater functional diversity of plant species including ferns (e.g. Blechnum chambersii, Cyathea medullaris, Pteridium esculentum), shrubs (Muehlenbeckia australis, Brachyglottis repanda) and juvenile individuals of tree species (e.g. Knightia excelsa, Melicytus ramiflorus). These native species are relatively widespread in New Zealand and do not represent rare or endangered taxa (Allan Herbarium 2000), however, they commonly dominate forest succession and thus play an important role in driving vegetation change (Wardle 1991).
The community shift for non-native species from managed to boundary zones was from widespread European pasture grasses and herbs to shrubs (see ordination results above). Many of these species are both widespread and abundant in New Zealand, and form the basis of many pastoral systems (e.g. the grasses Anthoxanthum odoratum, Dactylis glomerata, Holcus lanatus, and the N-fixing forb Trifolium repens). However, a few non-native species that are known to have large, negative impacts on native biodiversity or other ecosystem properties were present. For example, the clonal herb Tradescantia fluminensis was present in kiwifruit orchard boundaries, and has well documented negative impacts on litter-dwelling invertebrates, and the regeneration of indigenous tree species (Williams & West 2000; Williams et al. 2003; Standish et al. 2004). Similarly, native forests and ecotones are an important habitat for non-native weed species. For example, the shade-tolerant and high-impact invasive shrubs Lonicera japonica, Ligustrum lucidum, Solanum nigrum and Berberis glaucocarpa were observed only in forest and ecotone zones; although each of these species was observed in <10 plots (Appendix S3), all can persist in forest understories and are managed as environmental weeds. These and other non-native species are managed as environmental weeds to mitigate their negative impacts on indigenous biodiversity or alter ecosystem processes and services (Howell 2008; Peltzer et al. 2009, 2010).
There were few shared species between managed zones and those outside of management (i.e. ecotones and forests) for both native and non-native species. As a consequence, almost no weed species found within the managed zone were shared with unmanaged habitats (Appendix S3). An important exception to this was naturalized kiwifruit that was observed in two forest plots, suggesting kiwifruit production per se may not be a major source of new weeds into indigenous habitats. However, this observation may be because of successful weed control by regional government in order to prevent the spread of these weeds (Sullivan & Hutchison 2010), because time as naturalization has not been sufficiently long for these weeds to be abundant (Lockwood et al. 2005; Daehler 2009; Pyšek et al. 2009), or because there is an interaction between kiwifruit naturalization and dispersal agents like birds such that fruit size must first be sufficiently small in order to be eaten and dispersed (Logan & Xu 2006; J.J. Sullivan, 2011, pers. comm.). In general, kiwifruit orchards are not currently associated with increased numbers of invasive weed species at the spatial scale of our study (Tables 1, 2, Fig. 2; see also Pelosi & Goulard 2010).
If the land use of neighbouring properties is changed from grassland systems to farm forests, we would expect there to be a migration of woody weeds from forests into these systems. Although we did not determine the source of weed populations in forest gully systems, anecdotal evidence suggests that dumping of garden waste and seed dispersal by birds have both created nascent foci for non-native species. Similarly, Sullivan et al. (2004, 2005) found that the richness and abundance of environmental weeds is strongly associated with garden dumping and close proximity to urban areas. On the other hand, weed species in agroecosystems rarely also occurred in neighbouring ecotone or forest habitats, even though some weeds such as the grasses Holcus lanatus and Bromus diandrus, and the shrub Ulex europeaus persist at much lower abundance in forest canopy gaps (M. Perry, 2011, pers. comm.). In summary, our findings demonstrate that there are distinct communities of both native and non-native plant species between managed and unmanaged habitats. We also suggest that a two-way flow of species between managed and unmanaged systems is unlikely because of the coincident shift from grassland (managed) to forest (unmanaged) vegetation in this system.
In contrast to previous studies (Stohlgren et al. 2006; Fridley et al. 2007), native plant diversity was negatively associated with non-native plant diversity across all habitats for both kiwifruit and grassland systems. This negative association held across changes in plant growth-form and species composition associated with shifts from herbaceous and grass-dominated managed systems, to shrub-dominated ecotones, to tree-dominated native forests (ordination results for changes in plant community composition above; see also Appendices S1,S2). The negative relationship between native and non-native plant species richness and diversity across all management zones was surprising, and highly robust given the distinct native and non-native plant communities observed in managed and unmanaged habitats here. The cause and effect of this relationship cannot be determined using our community compositional data, but could be disentangled using appropriate manipulative field experiments or more complex spatial analyses (Peltzer et al. 2009; see also Pelosi & Goulard 2010; Ernoult & Alard 2011).
Land use comprises a suite of management activities, and these differ considerable between kiwifruit and grassland land uses. For example, grassland systems are seasonally grazed by domestic stock and receive relatively low inputs of fertilizer and or herbicide use whereas kiwifuit orchards are typically ungrazed and have moderate herbicide application (Steven & Benge 2007; Macleod et al. 2012). Similarly, although disturbance regimes increase in magnitude and frequency in managed systems, forests are not undisturbed, but rather have a history of canopy disturbance through windthrow, soil erosion (because of their location in gullies), introduced pest animals, and fire (Wardle 1991). Our results above thus demonstrate that the different subsets of native and weed species observed across habitat-types are driven by a complex set of mechanisms associated with different land uses (see also Macleod & Moller 2006; Lee et al. 2008). This further highlights the issue that land use and intensification comprises a complex suite of inputs and management activities. More generally, managed systems alter species composition, resource availability and disturbance regime, which interact to determine the balance between native and non-native species (Sandel & Corbin 2010; Laliberté et al. 2013; Tomasetto et al. 2013). The effects of land use and management per se on native and non-native plant species require additional information on resource manipulations and disturbance.