Ecological, taxonomic and life history correlates of seed mass among Indiana dune angiosperms.
The ecological and taxonomic correlates of seed mass variation were evaluated in 648 angiosperm species of the Indiana Dunes region. Each species occurred in ≥1 of 13 habitat types described in the Indiana Dunes flora. To determine the effect of apparent water and light availability on seed mass, each habitat was assigned to 1 of 4 categories representing combinations of inferred water and light availability. Life histories or life forms represented were: annuals, biennials, herbaceous vines, parasites, perennials, short-lived perennials, shrubs, small trees, trees and woody vines. Two phenological variables were available for most species: the time at which flowering began (early, middle or late) and the duration of flowering (short: <2 months; medium: 2-3 months; long: >3 months). The mean seed mass of each species was established by weighing samples from herbarium specimens in the United States National Herbarium. The frequency distribution of raw seed mass is highly skewed among species, so mean seed mass for each species was assigned to 1 of 14 seed mass classes based on a log scale. One-way ANOVAs revealed a significant effect of all ecological and taxonomic factors on seed mass, with the exception of native vs. alien status. Seed mass variance was accounted for as follows: family, 30%; life history, 22%; habitat, 8%; water/light category, 5%; onset of flowering, 5%; and duration of flowering, 4%. The smallest seeds were produced by the Scrophulariaceae and the largest by the Fabaceae, Liliaceae, and Rosaceae. The largest seeded species lived in close habitats: wooded dunes, thickets and wet dunes. Small-seeded species were associated with open habitats: wet dunes, aquatic habitats, marshes and streamsides. Light availability was a better predictor of relative seed mass than was inferred water availability. The strong effect of life history on seed mass was due primarily to the extremely large seeds produced by trees. Significant differences in seed mass were not detected among life forms producing seeds of intermediate mass (seeds produced by annuals, biennials, perennials, shrubs, small trees and vines were statistically indistinguishable). Early flowering and flowering of short duration were associated with the production of large seeds. Two-way ANOVAs detected 8 significant two-way interactions: family × native vs. alien status, family × life history, family × onset of flowering, habitat × life history, water/light category × life history, water/light category × native vs. alien status, life history × duration of flowering and onset of flowering × duration of flowering. These were further evaluated to determine the source of the interaction. Multi-fractional ANOVAs provided measurements of the effects of 1 class variable on seed mass while all other variables were controlled statistically. The statistical effect on seed mass of each variable independent of the others was much lower than in the one-way ANOVAs. Seed mass variance was explained independently by each variable as follows: family, 13%; life history, 13%; habitat, 4%; water/light category, 3%; onset of flowering, 1%; and duration of flowering, 1%. Two strong associations were detected in these ANOVAs: an association between family and habitat and between life history and onset of flowering. Although the effects of ecological and life history characters on seed mass evaluated in this study were highly statistically significant, the R2 values associated with these effects were quite low. For example, habitat accounted for only 4% of the variance in seed mass independent of the other effects measured. This suggests that at the geographical scale investigated in this study, there is no primary habitat-specific ecological attribute that determines the seed mass of a habitat's component species. Seed mass of a species is determined by a combination of its phylogeny, life history, phenology and ecological attributes, including variables were measured in this study. Even the most complete ANOVAs (including effects due to all variables measured) accounted for no >44% of the total variance in seed mass among species.