Summary1. Plant invasions are predicted to accelerate in a world with increased anthropogenic disturbance. Non-native species pre-adapted to these disturbances may especially be poised to invade novel communities. Conservation managers therefore need predictions of how to alter disturbances to maximize the persistence of native biodiversity. 2. We tested a multivariate hypothesis about the causal mechanisms underlying plant invasions in an ephemeral wetland in South Island, New Zealand, to inform management of this biodiverse but globally imperilled habitat. Our approach details among the first applications in ecology of Bayesian structural equation modelling, demonstrating its potential to inform management by disentangling the relative importance of strongly intercorrelated processes. 3. We found that invasion by non-native plants was lowest in sites where the physical disturbance caused by flooding was both intense and frequent. This effect was stronger than the positive response of non-native species to high soil N supply, which was positively related to flooding. 4. Sites flooded over a 4-year period had greater reductions in invasion than those associated with floods in the year prior to plot measurement because non-native species lacked traits for long-term persistence beneath water. Grazer exclusion had a small positive effect on invasion, as non-native species were preferentially selected by the herbivores at our site. 5. Our results show that only species adapted to the dominant disturbance regimes at a site may become successful invaders. Species native to ephemeral wetlands have specially evolved traits that allow them to persist and dominate in these sites. 6. Synthesis and applications. Predictions of invasions in a world of multiple disturbances clearly need to consider whether the evolutionary history of non-native species predisposes them to invade novel communities. Maintaining hydrological and nutrient regimes of ephemeral wetlands will limit the number of introduced species that are pre-adapted to become invasive.

Abstract:Estimates of vegetation attributes measured by sampling often inform scientific inference, management actions, and policy decisions. However, different sampling methods and sample sizes (i.e. number of plots) can yield significantly different estimates of vegetation attributes. This occurs because the abundance distributions and spatial distributions of species in the plant community influence their probabilities of detection and estimates of their abundances. We predicted that different sampling methods and sample sizes would produce significantly different estimates not only of vascular plant species diversity, but also of indigenous dominance (the level of indigenous influence) in mixed vegetation where indigenous and exotic floras have different abundance or spatial distributions. To test our predictions we applied three sampling methods to 24 plots in grassland and cushion vegetation in a 1058-ha scientific reserve in the Upper Waitaki (Mackenzie) Basin, New Zealand. Our methods sampled ground areas from 0.65 to 400 m 2 , and included two variants of common 'subsampling' approaches, which assessed only discrete subunits within larger plots. Indigenous plant species were both less abundant on average and more spatially-clustered (i.e. less evenly dispersed across plots) than exotic species. The two subsampling methods were less likely to detect less abundant and more spatially clustered species, leading to lower ratios of indigenous to exotic species recorded, and lower estimates of indigenous dominance of composition (% of species indigenous). Numbers of indigenous species accumulated more rapidly with increasing sample size than numbers of exotic species, so that indigenous dominance also increased with the number of plots sampled. We conclude that properly measuring species diversity and indigenous dominance in mixed indigenous-exotic plant communities requires both the searching of sizeable plots and use of rarefaction rather than plot-averaging of statistics. We suggest greater use of rarefaction and more consideration of species' detection probabilities in sampling New Zealand's mixed indigenous-exotic plant communities should improve the reliability, transparency and comparability of measures of diversity and may also provide new ecological insights.

Factors controlling vegetation restoration of depleted short-tussock grasslands are poorly understood. We investigated effects of mouse-ear hawkweed ('hawkweed', Pilosella officinarum) cover and environmental stress associated with landform and soil type on the rate and pattern of indigenous vegetation recovery from grazing in the highly-modified 1000-ha Lake Tekapo Scientific Reserve in the north of the Upper Waitaki ('Mackenzie') Basin. The reserve has been destocked of sheep and under effective rabbit control since 1992. At that time, mouse-ear hawkweed dominated vegetation on three of its five major landforms (up to 42% cover), and 44-89% of soil was exposed. In 2011 we resampled 12 original vegetation monitoring plots that were established in 1993. Indigenous vegetation recovered in the 18 years following removal from grazing despite high levels of initial modification and exotic cover dominance. Exposed soil and rock decreased, and indigenous plant cover and litter increased across all landforms, while vascular and nonvascular indigenous plant cover increased at different ends of a landform-driven productivity gradient. Hawkweed invasion did not retard recovery; to the contrary, the extent and rate of recovery was higher on more productive landforms with higher initial hawkweed cover. The pattern of change across the reserve was consistent with grazing having exerted a powerful constraint on the growth and biomass of both indigenous and exotic palatable species prior to reservation. Soil moisture stress appeared to delay the timing of hawkweed invasion, and to constrain productivity and hence the rate of indigenous vegetation recovery following release from grazing. We propose a testable model of interacting influences of grazing and environment on indigenous vegetation and the niche of mouse-ear hawkweed, in which recovery outcomes depend on environmental productivity. We suggest that removal of feral grazing as well as stock, size of recovery area, limited monitoring-disturbance, and timing of grazing-release relative to hawkweed invasion may explain why our results and conclusions contrast with those from other studies of release of fescue-tussock grasslands from pastoral grazing.