Seed dispersal is a fundamental process in plant ecology as it can drive diversity in a population and communities. Due to its high importance in the ecosystem, various modeling approaches have been developed to understand seed movement and dispersal mechanisms. Long-distance dispersal (LDD) plays a significant role in determining a population’s genetic diversity and factors influencing the occurrence of LDD are frequently explored. Animal-mediated seed dispersal is a key component of LDD events. Frugivores can travel long distances and transport seeds beyond the capacity of wind or other abiotic processes. Therefore, having good models for animal movement patterns becomes crucial in understanding the frequency and extent of LDD events. In this study we seek to understand how conventional animal movement models can be improved with the application of Extreme Value Theory (EVT) in statistics. Specifically, we test the use of a Generalized Pareto distribution to model animal movement. We also explore the underlying individual variation in frugivore movement patterns and its impact on the distance that seeds are dispersed. We apply these methods to empirical data sets that contain locations for the many-banded araçari, Pteroglossus pluricinctus, one of the primary frugivores for the Amazonian canopy tree, Virola flexuosa.
Results/Conclusions
We combined empirical data on animal movement, gut retention time (GRT) and tree crop size to simulate Virola seed dispersal. Radio-telemetry data on animal movement provided locations for 13 different individuals, which were then used to obtain movement rates and turning angle preference. We also included GRT from field trials in our simulations. The use of EVT, specifically the application of a generalized Pareto distribution, provides a better modeling tool for animal movement data. This finding is substantial, since most studies use conventional probability distributions to model movement bouts and can significantly underestimate the distances a frugivore can move. This approach also showed that there is significant variation in individual movement patterns which directly influences the estimated number of LDD events. The underlying variation in animal movement is strong enough that all tested distribution models showed a better approximation to the data when individual variation in movement was included. We found that models that ignore this underlying variation consistently underestimate the number of long-distance seed dispersal events. The use of this distribution could help dispersal ecologists better understand the dynamics between frugivore individual variation and the extent to which they disperse seeds.