<p>Introduced mammalian predators, namely possums, stoats and rats, are the leading cause of decline in native avifauna in New Zealand. The control of these species is essential to the persistence of native birds. A major component of mammal control in New Zealand is carried out through the aerial distribution of the toxin sodium monofluoroacetate (otherwise known as 1080). The use of this toxin, however, is subject to significant public debate. Many opponents of its use claim that forests will ‘fall silent’ following aerial operations, and that this is evidence of negative impacts on native bird communities. With the continued and likely increased use of this poison, monitoring the outcomes of such pest control operations is necessary to both address these concerns and inform conservation practice. The recent growth in autonomous recording units (ARUs) provides novel opportunities to conduct monitoring using bioacoustics. This thesis used bioacoustic techniques to monitor native bird species over three independent aerial 1080 operations in the Aorangi and Rimutaka Ranges of New Zealand. In Chapter 2, diurnal bird species were monitored for 10-12 weeks over two independent operations in treatment and non-treatment areas. At the community level, relative to non-treatment areas, the amount of birdsong recorded did not decrease significantly in treatment areas across either of the operations monitored. At the species level, one species, the introduced chaffinch (Fringilla coelebs), showed a significant decline in the prevalence of its calls in the treatment areas relative to non-treatment areas. This was observed over one of the two operations monitored. Collectively, these results suggest that diurnal native avifaunal communities do not ‘fall silent’ following aerial 1080 operations. The quantity of data produced by ARUs can demand labour-intensive manual analysis. Extracting data from recordings using automated detectors is a potential solution to this issue. The creation of such detectors, however, can be subjective, iterative, and time-consuming. In Chapter 3, a process for developing a parsimonious, template-based detector in an efficient, objective manner was developed. Applied to the creation of a detector for morepork (Ninox novaeseelandiae) calls, the method was highly successful as a directed means to achieve parsimony. An initial pool of 187 potential templates was reduced to 42 candidate templates. These were further refined to a 10-template detector capable of making 98.89% of the detections possible with all 42 templates in approximately a quarter of the processing time for the dataset tested. The detector developed had a high precision (0.939) and moderate sensitivity (0.399) with novel recordings, developed for the minimisation of false-positive errors in unsupervised monitoring of broad-scale population trends. In Chapter 4, this detector was applied to the short-term 10-12 week monitoring of morepork in treatment and non-treatment areas around three independent aerial 1080 operations; and to longer-term four year monitoring in two study areas, one receiving no 1080 treatment, and one receiving two 1080 treatments throughout monitoring. Morepork showed no significant difference in trends of calling prevalence across the three independent operations monitored. Longer-term, a significant quadratic effect of time since 1080 treatment was found, with calling prevalences predicted to increase for 3.5 years following treatment. Collectively, these results suggest a positive effect of aerial 1080 treatment on morepork populations in the lower North Island, and build on the small amount of existing literature regarding the short- and long-term response of this species to aerial 1080 operations.</p>
Dynamic Evolution Model of a Collaborative Innovation Network from the Resource Perspective and an Application Considering Different Government Behaviors
