Abstract
Seed dispersal allows plants to seek favourable conditions or spread the risk of unfavourable conditions through space and time. While theory predicts a trade-off between spatial and temporal dispersal, empirical tests have been stymied by the difficulty of measuring dispersal directly and limited data on seed traits beyond mass.Using hidden Markov models and 20 years of plant community data, we estimated spatial and temporal dispersal probabilities for 60 native and non-native annual plants and linked them to extensive seed trait measurements.We found a clear spatial–temporal dispersal trade-off: non-native grasses showed high spatial dispersal, native forbs high temporal dispersal and non-native forbs spanned the spectrum. Importantly, the traits underlying these strategies differed by group. For native forbs, spatial dispersal increased with large, elongated seeds, while small, round seeds favoured temporal dispersal. In non-native forbs, spatial dispersal declined with size, and temporal dispersal was linked to coat thickness and nitrogen content. Seed mass showed little association with either spatial or temporal dispersal.Our results provide empirical support for the spatial–temporal dispersal trade-off while revealing that multiple distinct trait combinations can achieve similar dispersal strategies, underscoring the need to move beyond seed mass to understand plant life-history strategies.