Restoration efforts often focus on changing the composition and structure of invaded plant communities, with two implicit assumptions: 1) functional interactions with species of other trophic levels, such as pollinators, will reassemble automatically when native plant diversity is restored; and 2) restored communities will be more resilient to future stressors. However, the impact of restoration activities on pollinator richness, plant-pollinator interaction network structure, and network robustness is incompletely understood. Leveraging a restoration chronosequence in Pacific Northwest prairies, we examined the effects of restoration-focused prescribed fire and native forb replanting on floral resources, pollinator visitation, and plant-pollinator network structure. We then simulated the effects of plant species loss/removal scenarios on secondary extinction cascades in the networks.  Specifically, we explored three management-relevant plant loss scenarios (removal of an abundant exotic forb, removal of an abundant forb designated a noxious weed, and loss of the rarest native forb) and compared them to control scenarios. Pyrodiversity, proportion of area recently burned, and cumulative replanting effort (plugging and seeding) over the prior 10 years increased the abundance and diversity of floral resources, with concomitant increases in pollinator visitation and diversity. Pyrodiversity also decreased network connectance and nestedness, increased modularity, and buffered networks against secondary extinction cascades.  Rare forbs contributed disproportionately to network robustness in less restored prairies, while removal of typical “problem” plants like exotic and noxious species had relatively small impacts on network robustness, particularly in prairies with a long history of restoration activities. Restoration actions aimed mainly at improving the diversity and abundance of pollinator-provisioning plants may also produce plant-pollinator networks with increased resilience to plant species losses.

We performed floral patch-focused observation at approximately the same seasonal time points as the floral density data were collected. For each forb species that flowered at a site, we selected an observation patch by choosing a random transect number to provide a general location in the field, then located the nearest floral patch for the target species where the number of observable flowering units exceeded a minimum of 20 within an estimated area of approximately 2 m². We used this threshold to reject very low-density forb species for which visual display was unlikely to attract pollinators. Observers performed 15-minute timed observations when temperatures were between 15º and 30º C, humidity was above 20%, and wind speed was less than 3 m/s. Observers captured and dispatched all insect visitors that contacted stamens or stigmas. Lepidoptera and queen bumble bees were excluded from captures and were identified on the flower, due to conservation considerations. Timing was paused during specimen handling. Two observations were performed for every forb species at a site at each time point, carried out by different observers and with different diurnal timing to the greatest degree possible. Specimens were later identified to the lowest taxonomic level possible by taxonomic experts (e.g., genus or species for most bees and flies).

Data files can be accessed using Excel.