1. The balance of pollination competition and facilitation amongst co-flowering plants and abiotic resource availability can modify plant species and individual reproduction. Floral resource succession and spatial heterogeneity modulate plant-pollinator interactions across ecological scales (individual plant, local assemblage, interaction network of agroecological infrastructure across the farm). Intraspecific variation in flowering phenology can modulate the precise level of spatio-temporal heterogeneity in floral resources, pollen donor density and pollinator interactions that a plant individual is exposed to, thereby affecting reproduction.

2. We tested how abiotic resources and multi-scale plant-pollinator interactions affected individual plant seed set, modulated by intraspecific variation in flowering phenology and spatio-temporal floral heterogeneity arising from agroecological infrastructure. We transplanted two focal insect-pollinated plant species (Cyanus segetum and Centaurea jacea, n = 288) into agroecological infrastructure (10 sown wildflower, 6 legume-grass strips) across a farm-scale experiment (125 ha).

3. We applied an individual-based phenologically explicit approach to match precisely the flowering period of plant individuals to the concomitant level of spatio-temporal heterogeneity in plant-pollinator interactions, potential pollen donors, floral resources and abiotic conditions (temperature, water, nitrogen).

4. Individual plant attractiveness, assemblage floral density and conspecific pollen donor density (C. jacea) improved seed set. Network linkage density increased focal species' seed set and modified the effect of local assemblage richness and abundance on C. segetum. Mutual dependence on pollinators in networks increased C. segetum seed set, while C. jacea seed set was greatest where both specialization on pollinators and mutual dependence was high. Abiotic conditions were of little or no importance to seed set.

5. Intra- and interspecific plant-pollinator interactions respond to spatio-temporal heterogeneity arising from agroecological management affecting wild plant species reproduction. The interplay of pollinator interactions within and between ecological scales affecting seed set implies a co-occurrence of pollinator-mediated facilitative and competitive interactions among plant species and individuals.

The data were collected in May-September 2019 on a farm-scale agroecological experiment (INRAE CA-SYS platform) in Burgundy, France (47°19'06.7"N 5°04'17.6"E), with the aim of studying the influences of multi-scale plant-pollinator interactions on wild plant reproduction in agroecological infrastructures. Each experimental unit in the dataset is an individual C. segetum (n = 144) or an individual C. jacea (n = 144) plant. All 288 focal plants were transplanted in 16 agroecological infrastructures across the farm (10 wildflower strips and 6 grass/legume strips).

Surveys on the field included:

* monthly focal plant development monitoring (counting of composite flowers, fruits,...);

* monthly botanical surveys : 6 quadrats of 2m x 50cm along a transect of 100m x 2m centered on the focal plot (n = 16 transects), identification of flowering plant species + counting of individuals and floral units per plant species;

* bimonthly surveys of plant-pollinator interactions on the focal plants (plot-scale surveys, 15 minutes, n = 16 plots) and in the local floral assemblage surrounding the focal plants (15 minutes of slaw walk along the 100m x 2m transect, n = 16 transects): capture of insect floral visitors (Hymenoptera, Diptera, Lepidoptera) feeding or entering in contact with the anthers/pistil. The insect captured were identified in the lab using reference keys.

An automated meteorological station located on the experimental farm also provided daily records of temperature, precipitation, wind, etc.

All focal plants were harvested at the end of the species' flowering period (C. segetum : mid-July; C. jacea : mid-September). The biomass of fruits, seeds and the whole plant was measured in the lab, and the floral heads (fruits with or without seeds) and seeds were counted.

We summed the data of plant-pollinator interactions on the 16 focal plots and in the 16 local floral assemblages to model weighted bipartite interaction networks matching the flowering period of each focal plant individual. Network indices were computed using the R package "bipartite" (linkage density, H2', focal species d', ...) or manually (weighted total mutual dependence).

All variables in the dataset were calculated from data matching exactly the flowering period of each focal plant individual.

Microsoft Excel