Predicting Plant Migration Rates in a Changing World: The Role of Long-Distance Dispersal

Models of plant migration based on estimates of biological parameters severely underestimate the rate of spread when compared to empirical estimates of plant migration rates. This is disturbing, since an ability to predict migration and colonization rates is needed for predicting how native species will distribute themselves in response to habitat loss and climate change and how rapidly invasive species will spread. Part of the problem is the difficulty of formally including rare long-distance dispersal events in spread models. In this article, we explore the process of making predictions about plant migration rates. In particular, we examine the links between data, statistical models, and ecological predictions. We fit mixtures of Weibull distributions to several dispersal data sets and show that statistical and biological criteria for selecting the most appropriate statistical model conflict. Fitting a two-component mixture model to the same data increases the spread-rate prediction by an average factor of 4.5. Data limit our ability to fit more components. Using simulations, we show that a small proportion (0.001) of seeds moving long-distances (1-10 km) can lead to an order of magnitude increase in predicted spread rate. The analysis also suggests that most existing data sets on dispersal will not resolve the problem; more effort needs to be devoted to collecting data on long-distance dispersal. Although dispersal had the strongest effect on the predicted spread rate, we showed that dispersal interacts strongly with plant life history, disturbance, and habitat loss in influencing the predicted rate of spread. The importance of these interactions means that an approach that integrates local and long-distance dispersal with plant life history, disturbance, and habitat availability is essential for predicting migration rates.