COS 40-1
MOVED TO COS 103-1 THURS PM 1:30 L100B // Estimating seed dispersal distances with incomplete genetic data: new methods, power analyses and a case study of the tropical tree Tabebuia rosea

Tuesday, August 6, 2013: 1:30 PM
L100D, Minneapolis Convention Center
Marta I. Vargas-Timchenko, Smithsonian Tropical Research Institute, Panama, Panama
Helene C. Muller-Landau, Smithsonian Tropical Research Institute, Balboa, Ancon, Panama
Kristin Saltonstall, Smithsonian Tropical Research Institute, Panama
Emily V. Moran, School of Natural Sciences, UC Merced, Merced, CA
F. Andrew Jones, Botany and Plant Pathology, Oregon State University, Panama City, OR
Background/Question/Methods

Genetic data linking seeds directly to parents through maternal seed tissue are often hailed as the best way to obtain information on seed dispersal distances.  However, DNA quality in maternally derived seed tissue is often low, leading to high rates of genotyping errors, and usually much discarding of data.  Our objective was to test and apply methods for gleaning information on seed dispersal distances from incomplete and error-prone genetic data, using the tropical tree Tabebuia rosea as a case study. Adult trees and dispersed seeds were genotyped using microsatellite markers. Genotyping error rates were calculated from replicated genotypes and then incorporated into a model (based on competing sources model, CSM) to estimate seed dispersal distances using all data, including both complete and incomplete genotypes. We evaluated the effects of both genotyping error rates and the number of seeds genotyped upon dispersal estimates using simulations. In addition, we used basic circular statistics to look for directional patters of seed dispersal for the species.

Results/Conclusions

Simulations showed that for our system, datasets as large as 1000 genotyped seeds are not sufficient to estimate true dispersal distances. Realistic levels of genotyping errors for molecular markers with the resolution of those used here decrease the information content, such that approximately twice as many seeds are needed to obtain the same precision.  Our results demonstrate the importance of calculating genotyping error rates, and the value of including incomplete genetic data in analyses in order to increase power and obtain better parameter estimates. The data set used to estimate the dispersal parameters for T. rosea in this study (181 genotyped seeds) proved to be insufficient for the model to provide precise estimates of seed dispersal parameters.  The genetic data did provide useful information on directionality, showing significant bias towards due South, consistent with the prevailing northerly winds during the seed dispersal season of the focal species.