Skip to main content
Dryad logo

Laying low: Rugged lowland rainforest preferred by feral cats in the Australian wet tropics

Citation

Bruce, Tom et al. (2022), Laying low: Rugged lowland rainforest preferred by feral cats in the Australian wet tropics, Dryad, Dataset, https://doi.org/10.5061/dryad.2v6wwpzq4

Abstract

Invasive mesopredators are responsible for the decline of many species of native mammals worldwide. Feral cats have been causally linked to multiple extinctions of Australian mammals since European colonisation. While feral cats are found throughout Australia, most research has been undertaken in arid habitats, thus there is a limited understanding of feral cat distribution, abundance, and ecology in Australian tropical rainforests. We carried out camera-trapping surveys at 108 locations across seven study sites, spanning 200 km in the Australian Wet Tropics.  Single-species occupancy analysis was implemented to investigate how environmental factors influence feral cat distribution. Feral cats were detected at a rate of 5.09 photographs/100 days, 11 times higher than previously recorded in the Australian Wet Tropics. The main environmental factors influencing feral cat occupancy were a positive association with terrain ruggedness, a negative association with elevation, and a higher affinity for rainforest than eucalypt forest. These findings were consistent with other studies on feral cat ecology but differed from similar surveys in Australia. Increasingly harsh and consistently wet weather conditions at higher elevations, and improved shelter in topographically complex habitats may drive cat preference for lowland rainforest. Feral cats were positively associated with roads, supporting the theory that roads facilitate access and colonisation of feral cats within more remote parts of the rainforest. Higher elevation rainforests with no roads could act as refugia for native prey species within the critical weight range. Regular monitoring of existing roads should be implemented to monitor feral cats, and new linear infrastructure should be limited to prevent encroachment into these areas. This is pertinent as climate change modelling suggests that habitats at higher elevations will become similar to lower elevations, potentially making the environment more suitable for feral cat populations.

Methods

General camera deployment

108 Camera-trap pairs were placed along main roads, four-wheel-drive tracks, and walking trails with one camera-trap facing into the road and another camera-trap positioned in the habitat, 50 m perpendicular from the road. To test the hypothesis feral cats would be more likely to use roads in tropical forests we placed a camera in the forest. A length of 50m was used to counteract potential spatial avoidance by feral cats of habitat features favoured by dingoes (Canis lupus dingo), the apex predator in Australia (Fancourt et al., 2019). We treated both cameras, on-road and off-road, as a single site for analyses due to their proximity to one another. Each camera-trap pair's planned location was spaced 2.2 km along the road; this distance exceeded the predicted home range of 1.16 km for feral cats in productive, low seasonality environments like rainforests and matched the home-range estimate of female feral cats in a montane rainforest in Hawaii (Bengsen et al., 2016; Smucker et al., 2000). Nineteen opportunistic camera-trap pairs were placed on walking trails and old roads that were not present on maps of the study areas. These opportunistic deployments were placed 500 m-1.98 km from the nearest camera-trap pair using the road. This deployment strategy resulted in an average distance of 1.81 km between camera-trap sites based on the distance to the nearest neighbouring camera-trap site using the road network. Due to uncertainty around cat home ranges in the study area, there is a potential for individual feral cats to be detected at multiple cameras. To be conservative, we interpret the occupancy results as the probability of site use rather than true occupancy, which is the same approach as other studies on feral cats in Australia and carnivore surveys generally (Doherty et al., 2021; MacKenzie et al., 2017; Rogan et al., 2019). Surveys were conducted between April 2019 and July 2020 for a minimum of six weeks per survey. We did not use baits or lures in front of the camera-traps as their use can influence species behaviour in an unknown way if it has not been evaluated previously, causing either a repellent, attractive, or neutral response (Mills et al., 2019; Rocha et al., 2016).

Camera placement and settings.

We followed the general guidelines for camera-trap deployment recommended by Meek et al. (2012). At each camera-trap site, a Bushnell Trophy Aggressor No-glow camera (Bushnell Outdoor Products, Kansas, USA) was placed facing the road within 200m of the planned point. Camera-traps were placed facing the road, perpendicular to the direction animals would travel, to maximise the likelihood of detecting our target species (Wang et al., 2019). Camera-traps were positioned at a height of 20-45cm from the ground, as this is the approximate height of the centre of mass for an adult feral cat (McGregor et al., 2015). Cameras were angled to be parallel with the terrain they were facing. Cameras were not orientated at a specific compass bearing as we were following the road to place cameras and prioritised angling the camera perpendicular to the animal's expected travel route. For each road camera location, a single No-glow Bushnell Natureview (Bushnell Outdoor Products, Kansas, USA) was placed 50 m from the road in the forest with a consistent and unobstructed field of view. Vegetation and debris were removed to a minimum distance of 4 m in front of the camera-trap where necessary to ensure a clear field of vision over the survey period. All camera-traps were programmed to take three images per trigger, with no delay between triggers (recovery time between consecutive photos = 0.62 s for Bushnell Aggressor, 0.7 s Bushnell Natureview). Due to equipment failure, four of the forest camera-traps at Mount Zero-Taravale Wildlife Sanctuary were replaced with Reconyx Hyperfire PC800 infrared cameras. While different camera-trap brands have different detection probabilities, the fact that feral cats were rarely detected off roads in our study makes it unlikely this difference affected our results (Palencia et al., 2022). The infrared flashes on the road camera-traps were set to high to ensure as much of the road as possible would be visible in the pictures, whilst forest infrared flashes were set to low to avoid overexposing the image. The remaining settings were all left at their factory defaults.

Predictor variables of feral cat occupancy

To investigate spatial factors that influence feral cat occupancy, we collated environmental and anthropogenic variables that are hypothesised to alter feral cat distribution and site use (Doherty et al., 2015a). We tested hypotheses related to terrain (elevation, terrain ruggedness index), habitat (habitat type, understory vegetation density), anthropogenic disturbance (distance from the nearest human population, forest integrity, habitat fragmentation), primary productivity (mean annual rainfall), prey populations (biomass index of prey species), invasive herbivores (invasive herbivore trapping rate), and altered fire regimes (fire regime intensity). For a complete list of the variables considered, how they were calculated, and the hypothesis they are linked to see Table S1. As there is little known about the spatial scale at which predictor variables affect feral cat occupancy, each model covariate value was averaged over a 2.2 km2 circular area around the camera point (Stobo-Wilson et al., 2020). The averaging of a spatial covariate was done to capture the variable at a scale relevant to the home range of a feral cat in rainforest habitats. This is pertinent as the effective area surveyed by the camera trap is only relative to the sensor's detection zone, not the wider area we averaged for the environmental covariates (MacKenzie et al., 2017). All continuous variables were checked for correlation before analyses. If any variable pairing resulted in a Spearman correlation coefficient >0.7, we only retained what we considered to be the most biologically relevant variable. The remaining variables were then scaled to improve model performance. To consider survey biases, we used observation level covariates in the analysis. We used the survey effort (the total number of days both cameras were actively trapping at each site for each occasion) to account for variable camera performance of both the road and off-road cameras at each camera-trap site influencing detection probability. 

Usage Notes

The data provided are as follows;

1. Feral_cat_Pres_abs_DRYAD_FINAL.csv - A presence-absence matrix of feral cat detections. 1 means detected, 0 not-detected, NA the camera site was not active for most of the occasion (<3 days). 

2.Site_Covs_DRYAD.csv -  The site covariates needed to run the analysis. All need to be scaled to be run in the analysis. 

3.Obs_Covs_DRYAD.csv - Survey level co-variates, scaled camera-trap effort per site per occasion. 

Software needed is R and the packages required are in the Rmarkdown file. 

Funding

Ecological Society of Australia

Skyrail Rainforest Foundation