Direct estimation of pollen dispersal distances from paternity type analyses is often not possible.  Therefore, recent emphases have been on the development of indirect methods relying on the assumed decay with spatial distance in a measure of the genetic structure of pollen pools sampled by spatially separated traps (TwoGener and KINDIST). However, one of the challenges to using these models in natural settings is the extent to which their performance is affected by variation inherent to the dispersal system.  Therefore, we conducted a study in order to estimate the distribution of effective pollen dispersal by hawkmoths (Hyles lineata and Sphinx vashti) and bumble bees (Bombus species) in two study sites of Aquilegia coerulea with contrasting plant spatial structure (Grand Canyon National Park and Cedar Breaks National Monument).  We used microsatellite markers to genotype 118 geo-referenced mothers and their associated progeny arrays (1827 total offspring). The dispersal curve for A. coerulea individuals (the dispersal kernel) was estimated for hawkmoth and bumble bee pollination by fitting normal, exponential, exponential-power, geometric, and bivariate student's t distributions to a measure of correlated paternity using KINDIST.
Results/Conclusions

Analysis of molecular variance (AMOVA) for the 118 maternal plants suggested a diverse population of breeding individuals (Φ_PT = 0.089, P = 0.03).  AMOVA of the pollen pool sampled by the maternal plants indicated that A. coerulea sampled a diverse pollen cloud overall (Φ_FT = 0.267, P = 0.01).  The exponential-power function provided the best fit to the data in terms of least square residual for hawkmoth and bumble bee pollination in both study sites.  Assuming exponential-power decay in correlated paternity as a function of pair-wise distance; hawkmoths effectively dispersed pollen two to five times farther than bumble bees (31.67± 3.87 versus 10.37 ± 5.66 meters, respectively; P < 0.05 for each study site).  Dispersal was shorter where A. coerulea was more abundant, densely and continuously distributed (16.2 versus 25.8 meters, P < 0.01, for Cedar Breaks and Grand Canyon, respectively).  The larger difference in pollen dispersal between the two study sites for bumblebees suggested that bumblebees were affected more by A. coerulea spatial structure than hawkmoths.  Our study demonstrated that indirect methods of estimating pollen mediated gene dispersal such as KINDIST are sufficient to quantify the sources of variation which affect pollen dispersal in natural settings.