An understanding of the processes governing natural afforestation over large spatial scales is vital for enhancing forest carbon sequestration. Models of tree species occurrence probability in non-forest vegetation could potentially identify the primary variables determining natural afforestation. However, inferring processes governing afforestation using tree species occurrence is potentially problematic, since it is impossible to know whether observed occurrences are due to recruitment or persistence of existing trees following disturbance. Plant functional traits have the potential to reveal the processes by which key environmental and land cover variables influence afforestation. We used 10,061 survey plots to identify the primary environmental and land cover variables influencing tree occurrence probability in non-forest vegetation in New Zealand. We also examined how these variables influenced diversity of functional traits linked to plant ecological strategy and dispersal ability. Mean annual temperature was the most important environmental predictor of tree occurrence. Local woody cover and distance to forest were the most important land cover variables. Relationships between these variables and ecological strategy traits revealed a trade-off between ability to compete for light and colonize sites that were marginal for tree occurrence. Biotically dispersed species occurred less frequently with declining temperature and local woody cover, suggesting that abiotic stress limited their establishment and that biotic dispersal did not increase ability to colonize non-woody vegetation. Functional diversity for ecological strategy traits declined with declining temperature and woody cover and increasing distance to forest. Functional diversity for dispersal traits showed the opposite trend. This suggests that low temperatures and woody cover and high distance to forest may limit tree species establishment through filtering on ecological strategy traits, but not on dispersal traits. This study shows that 'snapshot' survey plot data, combined with functional trait data, may reveal the processes driving tree species establishment in non-forest vegetation over large spatial scales.