Data for: Wild bees and landcover: bee species’ body size does not predict the scale of effect, but bee phenology predicts association with landcover type

Habitat is a key aspect of any species’ niche and can affect populations at multiple spatial scales. Basic ecology and effective conservation thus require understanding which habitats matter and at which scales. Yet, habitat studies are rarely scale-optimized and what determines the scale(s) at which populations are affected by surrounding habitat (the “scale of effect”) is poorly understood. In this study, we test the “mobility hypothesis,” which predicts that species with larger foraging ranges should have larger scales of effect. The mobility hypothesis is the most popular explanation of what determines species’ scales of effect but empirical support is mixed. We test the mobility hypothesis using wild bee species and, in doing so, also assess landscape-scale habitat associations of 84 bee species. We collected 30,376 specimens of 84 bee species from 165 sites in the northeastern USA and used linear models to determine landcover associations and scales of effect for each species. To test the mobility hypothesis, we asked whether scales of the effect varied with two mobility-related traits - body size or sociality, which are the strongest known predictors of bee foraging ranges. Controlling the false discovery rate at 5%, we found 193 significant species-landcover associations across 60 (of 84) species. Scales of effect ranged from 100 to 8000 m (mode = 200 m; median = 1000 m) and – counter to the mobility hypothesis – were not associated with body size or sociality. As a result, we argue that ecologists should reconsider making assumptions about species’ scales of effect and should instead explicitly measure scales of effect for their particular study organism and system. Considering the landcover associations themselves, we found these were broadly explained by phenology, with spring-flying bees being associated with forests and summer-flying bees being associated with more open, non-forested habitats.

Methods
This analysis in this paper used a dataset amalgamated from five previous studies, each of which collected bees by pan or vane trap. Following is a brief summary of each of these studies. Data from three of these studies were previously published, and the publications are noted below and in the "Related Works" section, while data from the other two is being published here for the first time. Any future analyses using these data should cite the original articles.

Dataset 1: Pinelands (Winfree et al. 2007)

This study was designed to ask about the effects of human land use on bee communities. The study region was the Pine Barrens of southern New Jersey. There were 44 sites placed along a human land use gradient. In this region, the natural land cover is predominantly forested ericaceous heath, and human land cover is predominantly agriculture (blueberry, cranberry) and suburbs.

The 44 sites were visited 2-5 times each in 2003, for a total of 167 site visits. At each site visit, 44 pan traps (plastic bowls painted with white or fluorescent blue or yellow) were placed along a 110-m transect and left for 8 hours between 07:00 and 17:00. Data were only collected on sunny or partly sunny days.

Of our 84 focal species, 59 were detected in this study, represented by 1470 specimens.

Dataset 2: Biotic homogenization (Harrison et al. 2018a, b)

This study was designed to assess the role of human land use in biotic homogenization across space and ecoregion. The study region was New Jersey, New York, and Pennsylvania. There were 36 sites in a nested block design, with three blocks of three sites each nested within four ecoregions. Each block had one site embedded in each of three dominant landcover types - agriculture, (sub)urban, and forest – while local habitat was standardized as mown grass.

Sites were visited 1-6 times per year, from spring to autumn, in the years 2013-2015. Across sites and years, there were a total of 377 site visits. At each site visit, 36 pan traps and two vane traps were set out and left for 24 hours.

Of our 84 focal species, 83 were detected in this study, represented by 11923 specimens.

Dataset 3: SWG (previously unpublished)

This study was designed to examine differences in bee abundance, diversity, and community composition among different habitat types in New Jersey. There were 37 sites visited a total of 79 times between March and September of 2016. Sites were haphazardly located across the state, with representation of many major landcover and habitat types, including different forest types, crops, sub/urbanization, and near wetlands. Each site was visited at least twice, once during the spring and once during the summer. At each site visit, 35 white, blue, and yellow pan traps were placed along a 50-m transect for 5-7 hours between 08:00 and 15:00.

Of our 84 focal species, 77 were detected in this study, represented by 4154 specimens.

Dataset 4: Forests 1 (Smith et al. 2021)

This study was designed to assess the effects of forest age, area, and fragmentation on bee communities. The study region was the Piedmont ecoregion of New Jersey. There were 32 sites, all embedded within forests but with varying landscape contexts.

Sites were visited 2-4 times in 2017 and 2018, except 5 sites that were only visited in 2017. At each site visit, 39 pan traps were placed in a 40 m x 100 m grid. In 2018, four vane traps were also placed. Traps were left for ca. 8 hours between 05:30 and 20:00.

Of our 84 focal species, 64 were detected in this study, represented by 10550 specimens.

Dataset 5: Forests 2 (Winfree et al 2014)

This study was designed to assess the effects of forest vs. non-forest “matrix” habitat on bee communities. The study region was the Piedmont region of New Jersey. There were 16 forest sites embedded within forest fragments of differing size, or within sub/urban or agricultural matrix.

Sites were visited 4 times each in April and May of 2006. On each site visit, an array of 39 white, yellow, and blue pan traps were placed and left for 4 hours.

Of our 84 focal species, 55 were detected in this study, represented by 2279 specimens.