Seed‐Parent Fecundity Distributions in Bee‐Pollinated Forage Legume Polycrosses

Abstract

Modeling expected fecundity distributions in bee‐pollinated forage legume polycrosses allows more accurate assessment of true inbreeding levels expected in breeding polycrosses. Presented is an empirically derived model for expected seed‐parent fecundity distributions for a given polycross size (N). In this study, N standardized seed‐parent fecundity frequencies {[Pfi − (1/N)]/(1/N)} were modeled on 169 polycrosses in four species—alfalfa (Medicago sativa L.), birdsfoot trefoil (Lotus corniculatus L.), kura clover (Trifolium ambiguum Bieb.), and red clover (Trifolium pratense L.)—using Weibull distributions (scale parameter k, shape parameter λ). On the basis of all polycrosses, it was determined that k could be predicted from λ using a fifth‐order polynomial. It was also determined that λ estimates were lognormally distributed (scale parameter ς = 0.154, shape parameter σ = 0.340, r2 = 0.991). Polycross λ estimates were not correlated with polycross size (N); however, λ estimates were linearly correlated with a polycross's median number of seeds per plant (λ = 1.46 × 10−4 [median seed per plant] + 0.934, r2 = 0.446, P < 0.0001). Linear model comparisons between alfalfa and red clover polycrosses showed that the two species had significantly different linear models (P = 0.032). The relationship between λ estimates and median seed per plant is hypothesized to be the result of seed production in stressful environments, resulting in lower median seed‐parent fecundity and greater deviation from random mating in the form of larger λ values. Polycross seed‐parent fecundity λ estimates were linearly related to pollen‐parent fecundity λ estimates (r2 = 0.609, P = 0.022).