Land-use legacies affect flower visitation network structure after forest restoration

Agricultural land use causes drastic changes to ecosystems, which persist after agricultural activity has stopped. One way to mitigate these impacts is through restoration of post-agricultural lands; however, the interplay between agricultural history and restoration remains poorly understood. This is particularly true for interactions among species. We investigated the effect of experimental restoration of longleaf pine (Pinus palustris Mill.) forests with differing land use histories on floral visitation networks. We found that restoration of open canopy conditions caused drastic increases in floral and floral visitor abundance and species richness. We found that after restoration, plots with no history of agriculture supported more specialized floral visitations and networks than in post-agricultural plots. These results illustrate large positive effects of forest restoration and flowering of understory plants and floral visitation, along with a persistent agricultural land-use legacy affecting the structure of floral visitation networks that is still evident 66 years after agricultural abandonment.

Methods
Study site and experimental design

This study was conducted on the Savannah River Site, an 80,000 ha National Environmental Research Park in Aiken and Barnwell Counties, South Carolina, USA. It is a United States Department of Energy site with forest management conducted by the USDA Forest Service. Like much of the Southeastern United States, uplands at the Savannah River Site were historically dominated by longleaf pine (Pinus palustris Mill.) savannahs characterized by widely spaced canopy trees with a highly diverse groundcover plant community. Throughout its range, the vast majority (97%) of the area of this ecosystem has been lost to fire suppression and land conversion to agriculture. By middle of the 20th century, the majority of the Savannah River Site was being used for agriculture. When the Savannah River Site was acquired by the federal government in 1951, approximately 6000 residents were relocated to surrounding areas, and all agricultural activity was abandoned. The Forest Service then planted loblolly (Pinus taeda L.) and longleaf pine plantations on agricultural fields for the next several decades. 

Using aerial photography taken in 1951, we located eight sites which had been agricultural fields when the Savannah River Site was acquired and that were adjacent to stands of remnant longleaf pine savannahs. Remnant stands are defined as areas which have no known history of agricultural cultivation (i.e., tillage). In 2010 we established two pairs of one ha plots at each site (four 100 m × 100 m plots at each site, 32 plots total). Paired plots spanned across the land-use boundary between post-agricultural and remnant stands. In 2011, we randomly selected one half of the plots located in each land used type within a site (16 plots total) to undergo restoration. We thinned the canopy of the restoration plots to 10 large pine trees per hectare and planted longleaf pine seedlings. To promote grasses and forbs in the understory, herbicides were applied once in 2012 to control hardwood resprouts. The remaining plots received no restoration thinning or planting. Prescribed burning was also conducted across all plots regardless of restoration or land-use history type with either one or two burns being conducted in each experimental site between 2011 and 2017.

Floral visitation network sampling

We assessed visitation to plants by insects three times in each plot during September and October of 2017. Along a 50 by 10 m transect near the center of each plot, we either collected or recorded the count and identity of all floral visitors that we observed making contact with the reproductive parts of flowers, excluding spiders, thrips, and ants (no non-insect pollinators were observed). The transect was broken into 10 5×10 m subplots, and we spent three minutes collecting and observing in each subplot, spending a maximum of 30 minutes total per transect per observation period. If a subplot contained no open flowers, we noted this and moved to the next subplot. On sampling days, we conducted one observation period in the morning between 8:00 am and 12:00 pm and another one in the afternoon between 12:00 pm and 4:00 pm. We also recorded the approximate area covered by the blooms of each plant species within the 50×10 m transect once each sampling day. We increased the accuracy of the floral estimates by recording only the flowers present in one five by 10 m subplot at a time. We only assessed floral-visiting insect activity on sunny or partly cloudy days when wind gusts did not exceed 6 m/s and temperature was above 16° C. We identified insects to the lowest taxonomic level possible using published taxonomic revisions. Insect specimens are deposited in the J. B. Wallis / R. E. Roughley Museum of Entomology at the University of Manitoba, Canada. Some insects (380 individuals, 11% of total observations) which were observed but not collected could not be identified to species or morphospecies because of insufficient information. These interactions were dropped from the dataset and were not included in any of the results of the associated manuscript.