Changing assembly rules during secondary succession: evidence for non-random patterns

Describing the rules of community assembly is a central topic of ecology. Studying successional processes through a trait-based null model approach can help to better understand the rules of community assembly.According to theoretical considerations, at the beginning of succession - after getting over the dispersal limitation stage - community composition is primarily shaped by environmental filters (generating functional convergence), while in later stages limiting similarity (generating functional divergence) will be dominant. However, empirical evidence does not clearly support theoretical expectations.Our aim was to detect the presence and changes of trait-based assembly processes during old-field succession based on twelve traits. Changes in vegetation composition were evaluated by a combination of time series and space-for-time substitution: conducting three resurveys of permanent plots on four old-field age-groups. The individual dispersion of traits was transformed into effect size (i.e. departure from null model expectation). The impact of time since abandonment on effect sizes was tested by generalized additive mixed effect models.We detected a non-random pattern for each trait in at least some part of the succession. Departure from randomness did not change significantly over time for six traits: seed mass, lateral spread and pollination type were divergent, while leaf size, generative height and length of flowering were convergent. Six traits had changing patterns along the succession. Four of them showed increasing divergence (e.g. dispersal type, LDMC), which supports our hypothesis. While two (SLA, life form) displayed increasing convergence, contrary to expectations.We confirmed the general hypothesis that convergence is predominant initially and that divergence can be detected later in succession for four traits. However, the large variation found in trait dispersion indicates that complex processes operate during succession.