Mapping the stray domestic cat (Felis catus) population in New Zealand: Species distribution modelling with a climate change scenario and implications for protected areas

Abstract

a b s t r a c tSpecies distribution models of stray cats were developed using two types of occurrence data: (i) a combined dataset of stray cats and cat colonies in Auckland and projected to the wider New Zealand area; and (ii) population density as an analogue for country-wide stray cat occurrence. These occurrence data, together with sets of environmental variables were used as input to the Maxent modelling tool to produce maps of suitability for the species. Environmental variables used in the models consist of current bioclimatic conditions, and a future climate scenario (RCP8.5 for year 2070 CCSM model). Commonly occurring bias in the modelling process due to latitude, the area for selecting background points in model evaluation, inherent spatial autocorrelation of occurrence points, and correlated bioclimatic variables were explicitly addressed. Results show that the North Island consistently provide more suitable areas for stray cats with increased suitability in a high emission climate change condition. Key protected areas at risk from the increased suitability to stray cats are also presented.

Research challenges and conservation implications for urban cat management in New Zealand

“…Unowned urban cats are more prevalent in areas with higher human population density Farnworth 2012, 2013;Aguilar et al 2015) and live at far higher densities than do unowned cats in rural environs (Langham and Porter 1991). Proximity to human environments and anthropogenic food sources are likely to provide unowned urban cats with the necessary resources to reproduce and survive in significant numbers.…”

Mapping the Potential Global Range of the Brown Marmorated Stink Bug, Halyomorpha halys, with Particular Reference to New Zealand

“…The resulting ensembled predictions, in the form of rasters, were processed for presentation in ArcMap 10.2. The predictive maps were reclassified into low, medium and high categories using equal distribution categories in order to clearly present the differences between current and future suitability and measure the area that changed between periods [43]. Commonly used thresholds to determine presence absence were not utilized because of the reported insensitivity of some thresholds to the lack of absence data as well as the reported variability between different thresholds [57,58].…”

“…Commonly used thresholds to determine presence absence were not utilized because of the reported insensitivity of some thresholds to the lack of absence data as well as the reported variability between different thresholds [57,58]. The centroid of the high category area for each projection was also determined and plotted to show the direction and magnitude of movement or shift of the highly suitable areas [32,43,44]. Territorial boundaries were overlaid on the suitability model to provide information for management purposes.…”

“…These have been found to be high performance algorithms with occurrence only data [45]. A worldwide distribution model was generated for each of the algorithms and projected to New Zealand for current and future conditions [32,43,44]. Each model was evaluated according to Area Under the Curve (AUC), Kappa and True Skills Statistic (TSS).…”

“…VIF values for the non-correlated variables ranged from 3.11 to 7.06 or below commonly used thresholds [42]. Species distribution modeling was run using the R package "dismo" using three high performing algorithms; Random Forest (RF), Support Vector Machine (SVM) and Maxent, using the default values for each algorithm [43,44]. These have been found to be high performance algorithms with occurrence only data [45].…”