Seasonal dynamics of flower production at a community scale
First, seasonal dynamics of flower production in this area were analyzed using pooled data of all plots in each term. The number of flowering species and the species diversity of floral compositions by Shannon-Wiener’s H’ (based on the total flower number of individual species) were compared among observation periods (early, middle, late) and between years (2017, 2018). In the analyses, we conducted the comparisons of all species, bee-visited species, and fly-visited species, separately. In order to quantify the seasonal dynamics of flowering species, furthermore, dissimilarity of floral compositions was compared among seasons (early, middle, late) and between years (2017, 2018) by non-metric multi-dimensional scaling (NMDS), based on the Chao’s similarity index. These analyses were performed using the Vegan package ver. 2.6-4 (Oksanen et al., 2022) and the MASS package in R. The effects of season and year on floral compositions were assessed by permutational multivariate analysis of variance (PERMANOVA) using the adonis function.
Second, seasonal trends in the flower production of individual species were analyzed at plot base by a generalized linear mixed model (GLMM) with a Poisson error distribution and a log-link function, using the package glmmTMB in R. In this analysis, we specifically targeted for bee-visited and fly-visited plant species because of the dominance of these types (see Results). In the GLMM, flower number of each species in each plot was a responding variable, season, year, and pollination type (bee-visited or fly-visited) were fixed effect variables, and plot and species were set as random effect variables. In order to compare the seasonal patterns of flower production between bee-visited and fly-visited species, an interaction term between season and pollination type was included in the GLMM.