Get reliable science

The increasing need to manage plant invasions has generated international interest in predicting the dispersal of invasive weeds, including the role of humans due to the influence of anthropogenic factors on invasions. Tradescantia fluminensis, an invasive weed, is spread in New Zealand only by vegetative fragments, with long-distance dispersal thought to be largely by human dumping of garden waste. Here we test whether the occurrence of T. fluminensis is predictable from physical (slope, altitude, temperature, vegetation type and cover and stream presence) and anthropogenic (road surface type, proximity to settlement and proxies for ease of dumping) variables, to measure the influence humans exert on its dispersal. Sampling of both physical and anthropogenic variables was carried out with presence/absence of T. fluminensis noted at 151 sites throughout the Marlborough Sounds in New Zealand. The best model included four anthropogenic and three physical variables, with the anthropogenic variables accounting for 70% of the deviance explained. K-fold cross-validation showed the model had a success rate of 78%, correctly predicting 118 out of 151 observations. This was significantly better than a null model based only on the overall fraction of sites with T. fluminensis. These results indicate that within the Marlborough Sounds region, anthropogenic factors play a strong role in the dispersal of T. fluminensis in ways that are amenable to predictive statistical modelling.