Predicting invasion dynamics of four alien Pinus species in a highly fragmented semi-arid shrubland in South Africa

This study explored the determinants of spread of four alien Pinus species and the ability of models to predict invasion dynamics in a complex fragmen ted landscape. The role of environmental factors, natural and anthropogenic disturbance in relation to invasion history was assessed for different sta ges in the invasion process using a Geographic Information System. Pines es caped from plantations over the past 30 years and spread into the natural s emi-arid shrubland (renosterveld). The pattern of spread was compared with a simulated random distribution using two different techniques, a standard logistic regression, and a new recursive modelling approach (Formal Inferen ce-based Recursive Modelling; FIRM). FIRM analysis improved the accuracy of predictions and revealed interactive effects of variables hidden by the lo gistic regression analysis. More than 80% of isolated pine individuals were found in 20% of the habitat classified as suitable by the models. Soil pH was the most important predictor for the distribution of isolated trees, wh ereas the establishment of dense pine stands was largely determined by fire history. Differences in invasive behaviour could be explained by species a ttributes such as limited dispersal for P. canariensis, and better drought- tolerance for P. halepensis. Sixty-five percent of the current pine distrib ution was accurately predicted by the spatial distribution of the first tre es to have invaded. Such models could be used to predict potential spread o f invasive plants and gain a better understanding of the main factors drivi ng the invasion process. However, the spread of invasive species in fragmen ted landscapes, strongly modified by human activities, is very complicated, and the spread remains difficult to predict in the long term. The dynamics of invasion are discussed in relation to changes in land use and disturban ce regime.