Ecological niche modelling of an invasive alien plant and its potential biological control agents
Biological control, biological invasions, Campuloclinium macrocephalum, ecological niche modeling, Pompom weed
Invasive alien plants are of concern in South Africa. Pompom weed (Campuloclinium macrocephalum) is currently invading the Grassland and Savannah biomes of South Africa and is likely to continue spreading in the southern African sub- region. Two possible biological control agents (Liothrips tractabilis and Cochylis campuloclinium) have been identified for control of pompom weed. We used ecological niche modelling to predict which areas in southern Africa are likely to be suitable for pompom weed and the two potential biological control agents. The overlap between areas predicted to be highly suitable for pompom weed and areas suitable for the biological control agents was assessed. Methods of reducing sampling bias in a data set used for calibrating models were also compared. Finally, the performance of models calibrated using only native range data, only invaded range data and both were also compared. Models indicate that pompom weed is likely to spread across a greater region of southern Africa than it currently occupies, with the Savannah and Grassland biomes being at greatest risk of invasion. Poor overlap was found between the areas predicted to be highly suitable for pompom weed and those areas predicted to be suitable for the biological control agents. However, models of the potential distribution of the biological control agents are interpreted with caution due to the very small sample size of the data set used to calibrate the models. Models calibrated using both native range and invaded range data were found to perform best whilst models calibrated using only native range data performed the worst. There was little difference found between models that were calibrated using spatially reduced (selecting only one record per 30 min grid cell) and randomly reduced (randomly selecting 50% of available records) biased data sets.