Collective dispersal leads to variance in fitness and maintains offspring size variation within marine populations

Scott C. Burgess, Robin E. Snyder, Barry Rountree. Collective dispersal leads to variance in fitness and maintains offspring size variation within marine populations. The American Naturalist (accepted)

Abstract: Variance in fitness is well known to influence the outcome of evolution but is rarely considered in the theory of marine reproductive strategies. In coastal environments, turbulent mesoscale eddies can collect larvae into ‘packets’ resulting in collective dispersal. Larvae in packets return to the coast or are lost offshore in groups, producing variance in fitness. Using a Markov process to calculate fixation probabilities for competing phenotypes, we examine the evolution of offspring size and spawning duration in species with benthic adults and pelagic offspring. The offspring size that provides mothers with the highest mean fitness also generates the greatest variance in fitness, but pairwise invasion plots show that bet-hedging strategies are not evolutionarily stable: maximizing expected fitness correctly predicts the unique evolutionarily stable strategy. Nonetheless, fixation can take a long time. We find that selection to increase spawning duration as a risk-avoidance strategy to reduce the negative impacts of stochastic recruitment success can allow multiple offspring sizes to coexist in a population for extended periods. This has two important consequences for offspring size: 1) coexistence occurs over a broader range of sizes and is longer when spawning duration is longer, because longer spawning durations reduce variation in fitness and increase the time to fixation, and 2) longer spawning durations can compensate for having a non-optimal size and even allow less optimal sizes to reach fixation. Collective dispersal and longer spawning durations could effectively maintain offspring size variation even in the absence of good and bad years or locations. Empirical comparisons of offspring size would, therefore, not always reflect environment-specific differences in the optimal size.