Abstract
New Zealand’s braided rivers are a globally unique ecosystem and provide breeding habitat for several threatened species of endemic bird. One of the main threats to these bird populations is predation by up to nine species of introduced mammals, primarily feral cats (Felis catus), ferrets (Mustela furo), stoats (M. erminea) and European hedgehogs (Erinaceus europaeus). Breeding on islands may offer vulnerable animals some protection from mammalian predation because water acts as a barrier to movement – the ‘island-effect’. However, not all islands are safe from mammalian predators. The aim of this thesis was to understand what importance braided-river islands have as refugia for braided-river birds. In particular, which of the interconnected factors that may determine predator presence on braided-river islands are the most important, and how predator presence on these islands may be affected by reduced river flow. To address this aim, I investigated first whether there was more evidence for an island-effect on New Zealand’s braided rivers. I then derived efficient detection techniques for the mammalian predators I expected to find, and determined which factors best described a braided-river island’s relative isolation. I then examined factors associated with predator distribution during the bird breeding season, and ascertained which predators were most commonly found on islands. Finally, I determined which island variables explained predator presence on my study islands, cross-validating my models to assess their reliability. This study was undertaken mostly on the Rangitata River, a large braided river in South Canterbury, with a relatively unmodified upper section and a heavily-vegetated lowland section. It has islands with a range of sizes, isolation levels and amounts of vegetation and prey.
First, to investigate the support for an island-effect on New Zealand’s braided rivers, I compared mainland and island nest success for three braided-river bird species on six rivers, using existing nest monitoring data. I found that island nests tended to have higher survival rates than mainland nests, but that this trend was not consistent between and/or within species, rivers and/or years, suggesting that other factors, such as the relative abundance of available primary prey and the presence of avian predators, affect the strength of an island-effect.
Because no best-practice methodology for detecting the main predators on braided rivers exists, I compared the efficiency of nine mammalian predator detection techniques at 19 sites along the banks of the Rangitata River. The most efficient detection techniques—detecting predator species on 90% of survey lines within 3–18 nights—included large tracking tunnels and hair tubes for feral cats, large tracking tunnels for hedgehogs, and WaxTags® for brushtail possums (Trichosurus vulpecula). Conversely, none of the techniques were particularly efficient at detecting stoats and Norway rats (Rattus norvegicus). These findings emphasised the need to use more than one technique to detect a species, and enabled me to make recommendations regarding the detection techniques that might be suitable for introduced mammalian predators in braided-river and other non-forest habitats.
I created an index for the relative isolation of braided river islands using principal components analysis. Most of the variance in the data was captured within two principal components. The first component was associated with flow variables, such as channel discharge, velocity and depth; the second was associated with an island’s distance to the mainland and the number of channels to cross. This indicated that island isolation could be described best using one or more variables from each group.
Between 2007 and 2010, I surveyed for mammalian predator presence and visitation frequency at 18 mainland sites and 58 islands on the Rangitata River, over an 18-night period. I simultaneously measured a range of habitat variables, including island characteristics believed to influence predator presence on islands. The predator composition was similar in the upper and lower river reaches, although the heavily vegetated nature of the lower river favoured possum, and possibly stoat and weasel (Mustela nivalis), presence there. I found fewer mammalian predator species on islands compared with the adjacent mainland, and I recorded predator presence on only 64% of islands, providing more evidence for an island-effect on braided rivers. Cats and Norway rats were the predator species most likely to be found on islands. Comparison with the mainland distribution of predators indicated that ferrets (the most-commonly detected mustelid), hedgehogs and possums may be less inclined to cross channels than these other two species.
Predator presence on Rangitata River islands was associated with island size, presence of vegetation and lagomorphs, and the distance to the mainland. Predator visitation rates were highest on islands with moderate levels of vegetation cover in plots and visitation rates increased with the proportion of plots containing lagomorph sign. The association of predator presence with channel discharge was less clear, although cat and ferret presence was limited to islands with a total discharge < 6 and < 3 m3/sec, respectively. Therefore, cats are likely to be the mammalian predator with the largest impact on braided river birds, given their relative abundance on the mainland, frequency of occurrence on islands, ability to cross channels, and propensity to kill adult birds.
Finally, I developed general and species-specific models to investigate the relative importance of island characteristics and to determine the likelihood of predator presence on bare and vegetated islands as island size and relative isolation increased. Cross validation techniques indicated model fits were moderate–good, although the Norway rat model was classed also as having poor fit, possibly because Norway rat distribution was temporally and/or spatially patchy. The models predicted that predators are more likely to be found on an island following a reduction in its relative isolation because the probability of predator presence increased with decreasing distance to the mainland and discharge thresholds. The models enabled recommendations to be made also for which islands may be the best to focus management efforts on. Maintaining islands 1–3.5 ha, > 20 m from the closest mainland and with a total channel discharge > 6 m3/sec, free from vegetation and lagomorphs should be of most benefit for breeding braided-river birds. Vegetation and lagomorph removal from islands and the wider area would especially benefit breeding birds on small braided rivers. Predator control programmes should focus also on large, vegetated islands, especially islands that are little isolated from the mainland or from bird-breeding islands.