Results
Over the 5 years of study, we observed 4,261 total interactions across 836 links between 267 species of animal visitors to 41 species of plants. Flower visitor species consisted of 109 Hymenoptera, 63 Diptera, 36 Coleoptera, 35 Hemiptera, 21 Lepidoptera, 3 Orthoptera, 1 Raphidioptera (Agulla sp.), and 1 hummingbird (Selasphorus platycercus ). Plant species consisted of 14 Asteraceae, 5 Rosaceae, 3 Fabaceae, 3 Orobanchaceae, 2 Ranunculaceae, and 1 of each Apiaceae, Boraginaceae, Campanulaceae, Crassulaceae, Gentianaceae, Geraniaceae, Hydrophyllaceae, Melanthiaceae, Onagraceae, Primulaceae, and Rubiaceae. Removal of flower visitor groups not commonly regarded to be pollinators (Hemiptera, Orthoptera, Raphidioptera; but see (Wardhaugh 2015)) before analysis yielded nearly identical results The aggregated network showed a nested structure (NODF = 25) and connectance values (0.08) that are typical of plant–pollinator networks (Schwarz et al. 2020). The network had one module. Observed sampling completeness of interaction richness was at 52% of the Chao2 estimator, 64% of the Jack1 estimator, and 81% of the bootstrapped estimator. Observed sampling completeness of pollinator richness was at 62% of the Chao2 estimator, 71% of the Jack1 estimator, and 85% of the bootstrapped estimator. Observed sampling completeness of plant richness was at 90% of the Chao2 estimator, 93% of the Jack1 estimator, and 97% of the bootstrapped estimator. Environmental conditions, such as temperature, varied temporally throughout the season and across years (Fig S1), and spatially among plots, as did soil moisture (Fig S2).
Interactions between generalists showed higher temporal and spatial persistence. Interactions in the core of the nested network tended to have higher inter-annual persistence (Spearman’s rank correlation coefficient = 0.35, P << 0.001; Fig 1a), intra-annual persistence (Spearman’s rank correlation coefficient = 0.22, P << 0.001; Fig 1b), and inter-plot persistence (Spearman’s rank correlation coefficient = 0.35, P << 0.001; Fig 1c). Observed Spearman’s rank correlation coefficients for all three persistence variables were higher than those expected under null models (Fig S3).
Temporal and spatial interaction persistence values were interrelated. That is, interactions with higher interannual persistence tended to have longer phenophases (Spearman’s rank correlation coefficient: 0.59, P << 0.001) and be more widespread among plots (Spearman’s rank correlation coefficient: 0.73, P << 0.001). Interactions with longer phenophases tended to be more widespread among plots (Spearman’s rank correlation coefficient: 0.78, P << 0.001).
At the species level, inter-annual persistence, longer phenophases, and higher plot occurrence were associated with generalization (Fig 2, S4). For both plants and pollinators, species that had higher inter-annual persistence were closer to the core of the nested network (Fig 2a, b; for plants: R2 = 0.73, P << 0.001, for pollinators R2 = 0.61, P << 0.001). Both plant and pollinator species with longer phenophases were closer to the core of the nested network (Fig 2a, b; for plants: R2 = 0.72, P << 0.001, for pollinators R2 = 0.70, P << 0.001). Both plant and pollinator species that were more widespread among plots were closer to the core of the nested network (Fig 2c, d; for plants: R2 = 0.58, P << 0.001, for pollinators R2 = 0.57, P << 0.001). Using degree (number of links), as a proxy for generalization showed similar patterns as species proximity to the network core (Fig S4). Finally, phenophase length for plants and pollinators was associated with broader ranges of temperatures on the mornings of sampling days (for plants: Pearson’s r = 0.67; for pollinators Pearson’s r = 0.81).