Using functional dissimilarity among species pools and communities to predict establishment of native and alien species.
Aims: Predicting plant establishment and growth is a fundamental goal of community ecology, especially when studying invasion. Functional traits can infer a species' environmental tolerances and competitive abilities and may help improve predictions, yet such models are limited by a lack of data on failed establishment. Filtering of species from regional pools to site-specific pools to local communities provides abundant evidence of failed establishment represented by species excluded at each level of the hierarchy. We tested whether trait differences between observed and absent species among these pools can help predict species establishment. Location: Southern Estonia. Methods: For 31 semi-natural grassland sites, we collected plant composition data at multiple spatial scales to estimate the regional species pool, site-specific pool, and local community. We compared traits among these pools to model community assembly and predict which species could join the site-specific pool and local community. We tested these models using an experiment in which we added seed and transplants of 15 species (9 native and 6 alien) to intact and disturbed plots at each site. We considered differences in plant performance between intact and disturbed sites to be indicative of competition. Results: For all performance metrics, we found significant positive relationships with model predictions. These relationships were stronger for seed establishment than growth and were generally weak when predicting competition. Seed establishment was also best predicted in sites with high local diversity and low root biomass, potentially because of more accurate estimates of species pool and community composition. Seed establishment was better predicted for alien than native species and, although predictive power remained low, the opposite was true for competitive effects on establishment. Conclusions: Multi-scale models of community assembly may predict future establishment but must include other characteristics within such models. These models may perform poorly, however, when predicting biotic interactions.