An empirical framework for predicting and detecting long-distance larval dispersal events

Background/Question/Methods

Rare long-distance dispersal (LDD) is consequential to ecological processes such as colonization, range expansion, and the maintenance of connectivity across increasingly fragmented landscapes and seascapes. In systems where LDD is undertaken by small propagules in stochastic environments, such as larvae in the ocean, LDD is notoriously hard to predict and detect. Here, we develop an empirical framework for generating and testing LDD predictions by combining two major genetic methods: parentage analysis and population assignment tests. We hypothesize that dispersal kernels estimated from parentage analysis at relatively small spatial scales can accurately describe dispersal at larger spatial scales. We propose using such dispersal kernels to predict the frequency of LDD events between populations, then using population assignment tests to uncover LDD events and test those predictions. We applied this framework to a coral reef fish (Elacatinus lori) using a high-resolution data set of 931 adults and 3,828 recruits sampled across hundreds of kilometers of reef and spanning two generations.

Results/Conclusions

The parentage-based dispersal kernel predicted an exceptionally small number of LDD events: just 0-0.2% LDD in our E. lori samples. The population assignment tests revealed 0.1% of individuals were long-distance dispersers, consistent with predictions. These results support prior estimates of the tail of the species’ dispersal kernel. More broadly, these results demonstrate the usefulness of this new hypothesis-testing framework and reveal that even in stochastic marine environments, LDD can be accurately predicted. Therefore, we suggest that using complementary genetic approaches in a hypothesis-testing framework to explore dispersal at different spatial scales will facilitate a more complete understanding of larval dispersal patterns.