Applied statistical modelling to guide the control of introduced mammalian predators in the Murchison Mountains (Fiordland, New Zealand)

Abstract

Since the rediscovery of Porphyrio hochstetteri (takahē), in the Murchison Mountains (Fiordland, New Zealand) in 1949, there have been continued efforts to protect the species. Takahē numbers have been increasing from a minimum of 120 individuals in 1981, peaking in 2006 when 297 takahē were recorded (168 in the Murchison Mountains). Between 2002 and 2008, an experimental trapping programme was implemented within the Murchison Mountains targeting Mustela erminea (Linnaeus, 1758) (stoats) and Rattus rattus (Linnaeus, 1758) (ship/black rats). Approximately 40% of the Murchison Mountains was assigned for intensified predator trapping while the remaining area was left as a control. However, during the period of intensified trapping a stoat plague of unexpected magnitude occurred within the Special Takahē Area (STA); nearly half of the remaining takahē in the Murchison Mountains were killed. Thus, after introducing predator control in the area, the highest levels of observed predation occurred. In the previous decades of takahē monitoring and management with no stoat control, such high predation (40%) was never recorded. My MSc research primarily aims to elucidate the mechanisms underlying the 2007-2008 stoat plague and develop improved methods and tools for conservation managers. To facilitate the practical application of predictive models for management purposes, I devolop two fully featured software packages with a graphic user interfaces (GUIs) enabling quick, customisable approach to 1) generating artificial neural networks, and 2) a GIS platform to apply neural networks to an area. The predictive models I generate, combined with the software and improved CPUE metrics, can be used to optimise predator management in the Murchison Mountains.