Impacts of flowering density on pollen dispersal and gametic diversity are scale dependent

Pollen dispersal is a key evolutionary and ecological process, but the degree to which variation in the density of concurrently flowering conspecific plants (i.e., co-flowering density) shapes pollination patterns remains understudied. We monitored co-flowering density and corresponding pollination patterns of the insect-pollinated palm Oenocarpus bataua in northwestern Ecuador and found that the influence of co-flowering density on these patterns was scale-dependent: high neighborhood densities were associated with reductions in pollen dispersal distance and gametic diversity of progeny arrays, whereas we observed the opposite pattern at the landscape scale. In addition, neighborhood co-flowering density also impacted forward pollen dispersal kernel parameters, suggesting that low neighborhood densities encourage pollen movement and may promote gene flow and genetic diversity. Our work reveals how co-flowering density at different spatial scales influences pollen movement, which in turn informs our broader understanding of the mechanisms underlying patterns of genetic diversity and gene flow within populations of plants.

Methods
         We recorded monthly phenological state for all adults (n=181) within a core 130-ha plot by visiting each tree within the study plot and recording the number of reproductive structures. In 2015, we surveyed a 250 m buffer zone around the 130-ha plot and recorded 60 additional adults, all mapped and genotyped, but for which phenology was not recorded. This yielded 241 geo-located and genotyped candidate fathers, of which 181 also had monthly phenology data for analysis. To collect progeny for genetic analysis, we collected ripe fruits directly from the infructescences of O. bataua individuals; we refer to these trees as ‘maternal’ trees and pollen sources as ‘paternal’ trees, noting that the same individual could be both. We randomly sampled maternal trees with complete phenology records. We germinated the seeds in a nursery and collected and stored a tissue sample from the first leaf of each seedling. For each maternal tree, we used monthly phenology data to calculate the co-flowering density in the time frame that the maternal tree was flowering. We calculated density at two spatial scales: the ‘landscape’ scale, defined as the entire 130-ha study plot, and ‘neighborhood’ scale, defined as the average effective pollination neighborhood (Aep) of all progeny arrays included in this study, which was a circular area with a radius of ~ 320 m and an area of 33.40 ha. If the boundary of the neighborhood area extended beyond that of the study plot, neighborhood co-flowering density was estimated using the subset of individuals within the neighborhood area for which phenology data were available.

       We extracted genomic DNA from 962 offspring leaf samples representing 43 progeny arrays collected from 35 maternal trees. All samples were genotyped using 11 microsatellite loci through Polymerase Chain Reaction, following established protocols. We genotyped the 181 adults (pollen sources and maternal trees) in the study plot and the 60 additional adults in the 250 m buffer around the study plot using equivalent methods, totaling 241 genotyped parental trees. We used offspring and adult genotypes in the program CERVUS v. 3.0.3 (Marshall et al., 1998) to assign paternity with the aim of (1) calculating the average Aep, (2) quantifying the ‘observed’ distance pollen dispersed from paternal sources to maternal trees, and (3) estimating the paternal contribution to the genetic diversity of each progeny array. We used critical trio (Δ) values with at least 80% confidence and the following simulation parameters used previously for parentage analysis in our study area.