1. Cultivation of bioenergy feedstocks is a growing land-use worldwide, yet we have a poor understanding of how bioenergy crop management practices affect biodiversity. This knowledge gap is particularly acute for candidate cellulosic bioenergy feedstocks, such as tree plantations, and for organisms that provide important ecosystem services, such as pollinators.

2. We examined bee communities in 83 sites across three states in the southeastern USA—Alabama, Florida and Georgia. We compared bee abundance and diversity in 66 pine plantation sites that reflect management with and without potential bioenergy feedstock production. At least three bioenergy feedstock production methods have been proposed for this region: 1) converting conventional timber stands to short-rotation bioenergy plantations; 2) harvesting feedstock by thinning conventional plantations; and 3) harvesting of woody debris residues after plantations have been clear-cut.

3. We found that bioenergy-associated management practices including younger plantations (relative to older) and woody debris removal (relative to debris unremoved) in clear-cut plantations were associated with reduced bee diversity. Removing ground debris in clear-cut plantations also drastically increased bee abundance, though this effect was largely driven by strong dominance of just two bee species. Clear-cut plantations had lower beta diversity than standing plantations.

4. Synthesis and applications. Management practices associated with bioenergy feedstock production can have negative effects on bee community diversity. In particular, harvesting of debris in clear-cut plantations dramatically reduces bee diversity. Large-scale bioenergy feedstock production that increases the prevalence of young and clear-cut stands may cause landscape-level beta diversity to decline. Nevertheless, bioenergy pine plantations likely support higher bee diversity than corn fields, an alternative bioenergy feedstock.

Study sites and strata

We sampled bee communities from 83 sites, including 66 pine plantations, 10 natural reference-condition sites (longleaf pine forest remnants) and 7 corn production sites, distributed across the U.S. states of Florida, Georgia and Alabama, which are expected to be key bioenergy states. The sites were the same as those sampled for birds in prior work (Gottlieb et al. 2017) and clustered into three geographic ‘strata’ that did not follow state lines (Table 1). In pine plantations, we focus on three key attributes that reflect potential management changes for bioenergy feedstock production: 1) younger plantation age; 2) plantation thinning; and 3) harvesting of ground debris after plantations are clear-cut. We examined the effect of plantation age on bee communities by comparing young, unthinned plantations 8-12 years old (simulating harvest-ready bioenergy plantations) to more mature plantations 24-25 years old that have already been thinned. We also compared bee communities in plantations soon after thinning (simulating harvesting for bioenergy feedstock) to unthinned plantations of similar age (12-16 years). Finally, we examined the impact of harvesting plantation ground debris by comparing recently clear-cut plantations (felled within the last two years) with and without debris harvest. Each of the six ‘plantation types’ was represented in 9-12 sites (Table 1); all sites were >16 ha and spaced at least 2.5 km apart.

Table 1 Site replications per stratum for the six plantation types with and without potential management changes for bioenergy production, as well as natural longleaf pine stands and corn fields. Stratum S1 consisted of sites in Alabama; S2 in southern Georgia and the Florida panhandle; and S3 in north-central Florida.

Measuring local bee and plant communities

We surveyed bee communities over the spring and summer seasons of 2013, 2014 and 2015. Sites were not sampled repeatedly across years. In each site, we marked out two 2 x 200 m sampling transects at least 50 m from the plantation edge and at least 100 m away from one another.

For each sampling day, we collected bees using pan traps and aerial netting, which work effectively in tandem (Westphal et al. 2008). Pan traps consisted of small, plastic cups (3.25 oz., model P325, SOLO Cup Company, Illinois USA) painted with ultraviolet-bright blue, white or yellow paint, and filled with a dilute detergent-water solution that drowns the bees (Kearns & Inouye 1993; Westphal et al. 2008). Fifteen pan traps were held approximately 40 cm above the ground on wire stakes (VIGORO plant props, model 611872, Spectrum Brands Holdings Inc., Wisconsin USA, bent to better hold traps) so as to be visible above herbaceous vegetation, and positioned in alternating colors evenly along the center 100 m of the sampled transect. Pan traps were set up in the morning and collected after 24 hours. During each sampling day, we also performed targeted aerial netting of bees along the entire length of a transect for 30 minutes, excluding handling time for every successful capture with a stopwatch. Bee surveys were postponed on cloudy or rainy days, and each transect was sampled up to four times on separate days, amounting to up to 8 sampling days per site. Due to unforeseen weather and logistical difficulties, however, 9 of the 66 sites are represented by only one transect or by fewer than 3 samples. These sites were spread across two of the three strata and across land use categories. Still, we employed statistical methods robust to imbalances in sampling effort across our analyses. Collected bees were preserved in ethanol or pinned, brought back to the lab and identified to species or occasionally genus.

We measured available floral resources at each site once during the study on a sampling day, within 1 m of the central 100 m of each transect, by counting and identifying to species all understory non-grass plants that were in bloom. We surveyed floral resources among sites within a stratum as close together in time as possible to maximize comparability. A separate analysis of pollen loads present on some of the bee specimens from this work is covered in Bell et al. (2017), which describes interactions between bee and plant species.

Bee specimen preparation and identification

We usually pinned bee specimens on the day of sampling. We keyed out all specimens to genus, and 92% of taxa to species, using Discover Life online keys (https://www.discoverlife.org), in conjunction with Michener (2000) and Michener et al. (1994). Our identifications were verified by Ismael Hinojosa (Universidad Autonoma de Mexico) and Sam Droege (USGS); Sam Droege identified many specimens (particularly Lasioglossum) to species.