Modelling the distribution and relative abundance of feral camels in arid Australia

Abstract:

Determining the distribution and abundance of pest species is required for planning, directing and evaluating their control. This is particularly true for feral camels, which occur at relatively low densities in remote areas, and so control programs incur high travel costs. The broad-scale pattern of distribution of feral camels is known from infrequent aerial surveys that cover only part of the camel’s range. Estimates of abundance have therefore required extrapolation over both time and space (Saalfeld and Edwards 2010). A further problem is that the low density of camels combined with the low sampling intensity of surveys leads to imprecise estimates of density at a relatively fine scale (e.g. < 10,000 km), leading to potentially misleading distribution patterns that may result in misdirected management effort. Distribution patterns may be better represented with a spatial model that links aspects of the environment with probability of occupancy by camels and, ideally, density. Features of the environment attractive to camels can also be identified A resource selection function was fitted to data from a 2001 survey in the southern Northern Territory (McLeod and Pople 2010). This involved two steps. First, habitat suitability was modelled using habitat covariates for ~500 locations of camel groups and an equal number of ‘pseudo-absence’ locations selected randomly along transect lines where camels were not observed. Second, the relationship between density and habitat was modelled using a generalised additive model, conditional on camels being present. Habitat covariates included aspects of climate, distance to water sources, roads and human population centres, and the topography and broad vegetation class in 1 km and 5 km buffers around each location. The most parsimonious model identified a handful of high density ‘hotspots’. However, the model is static. If, as expected, camel population size continues to increase, the pattern of distribution may also change. A comparison of historic and future surveys can address this. Rainfall may also alter the pattern of distribution, as the high mobility of camels is well known. Rainfall or the normalised difference vegetation index could therefore be a useful predictor More recent survey data across all states containing feral camels now allows a more complete assessment of their habitat associations. Surveys are still spatially incomplete. A habitat model may best estimate the full distribution and indicate the Australia-wide population size.