Forest Ecology and Management 482 (2021) 118903

Contents lists available at ScienceDirect

Forest Ecology and Management

journal homepage: www.elsevier.com/locate/foreco

Bees in the trees: Diverse spring fauna in temperate forest edge canopies

Katherine R. Urban-Mead a,*, Paige Muniz˜ a, Jessica Gillung b, Anna Espinoza a, Rachel Fordyce a, Maria van Dyke a, Scott H. McArt a, Bryan N. Danforth a a Cornell Department of Entomology, Ithaca, NY 14853, United States b Department of Natural Resource Sciences at McGill University in Quebec, Canada

ARTICLE INFO ABSTRACT

Keywords: Temperate hardwood deciduous forest is the dominant landcover in the Northeastern US, yet its canopy is usually Temperate deciduous forests ignored as pollinator habitat due to the abundance of wind-pollinated trees. We describe the vertical stratifi­ Wild bees cation of spring bee communities in this habitat and explore associations with bee traits, canopy cover, and Forest management coarse woody debris. For three years, we sampled second-growth woodlots and apple orchard-adjacent forest Canopy ecology sites from late March to early June every 7–10 days with paired sets of tri-colored pan traps in the canopy (20–25 m above ground) and understory (<1m). Roughly one fifthof the known New York state bee fauna were caught at each height, and 90 of 417 species overall, with many species shared across the strata. We found equal species richness, higher diversity, and a much higher proportion of female bees in the canopy compared to the understory. Female solitary, social, soil- and wood-nesting bees were all abundant in the canopy while soil- nesting and solitary bees of both sexes dominated the understory. Canopy cover increased with leaf-out, and was negatively associated with understory but not canopy bee abundance. Site-level volume of coarse woody debris (CWD) did not predict bee abundance, while approximated landscape-scale availability of woody debris was positively correlated with understory wood-nesting and solitary-bee abundance. This work expands our understanding of habitats where bees are likely foraging and reveals vertically stratifiedbehavior. We emphasize deciduous forests as an important habitat for wild bee conservation and recommend further research into the behavior and diets of bees occupying the canopy, speculating that females forage for anemophilous tree pollen. Forest management plans that conserve above-ground deadwood may provide nest sites for wood-nesting bees.

1. Introduction resources, there is considerable anecdotal evidence that bees collect such pollen (reviewed in Saunders, 2018), and recent work further Studies of in both tropical and temperate forests emphasizes the importance of trees for resins, honeydew, and other non- frequently explore stratification, or distinct community associations floral resources (Requier and Leonhardt 2020). These canopies flower found at different heights above ground. Vertically partitioned com­ prolificallyin the early spring. Considering these possible resources and munities are often governed by reliance on tropical epiphyte habitats, unique habitat, we expect then that bees may be vertically stratified in (Ozanne et al. 2003), features such as soil moisture and wind protection the temperate hardwood canopy. (oribatid mites, Prinzing and Woas 2003), food resources (ants, Dejean Canopy and understory hymenopteran communities have been and Corbara, 2003; Basset et al., 2003), or a combination thereof (Halaj found to differ in some ways, with sometimes opposing trends (Ulyshen et al., 1998; Horvath´ et al., 2005; Klimes et al., 2015). Unlike many well- 2011). A study of predatory wasps with vane traps in Georgia, USA, studied tropical forests, temperate deciduous hardwood forests lack found that 99% of specimens were captured in the canopy (>15 m) features such as rich epiphyte communities, lianas, or suspended canopy compared to the understory (Ulyshen et al. 2010), while in Ohio, USA, soil (Chapin and Smith 2019). Flying may therefore have less parasitic wasps were more abundant in the understory than in yellow reason to distinguish or strongly associate with a given canopy layer in traps at 10 m (Pucci 2008). A season-long study in Georgia, USA found temperate ecosystems (Prinzing and Woas 2003). And although not only that the shiny green sweat bee Augochlora pura (Halictidae) was temperate wind-pollinated species are usually discounted as bee forty times more abundant in the canopy than the understory, but also

* Corresponding author. E-mail address: [email protected] (K.R. Urban-Mead). https://doi.org/10.1016/j.foreco.2020.118903 Received 27 October 2020; Received in revised form 24 December 2020; Accepted 26 December 2020 0378-1127/© 2020 Elsevier B.V. All rights reserved. K.R. Urban-Mead et al. Forest Ecology and Management 482 (2021) 118903 comprised 92% of all captured bees (Ulyshen et al. 2010). A similar that canopy floral resources may have important demographic impli­ result was found in North Carolina, USA, where understory and mid- cations for wild bees. We propose that a deeper understanding of the story (~9m) traps contained predominantly A. pura and Lasioglossum vertical dimension of pollinators in temperate forests will allow for their spp., with A. pura driving the higher mid-story abundances (Campbell explicit incorporation in management plans. et al. 2018). In a more recent study with both mid-story (5 m) and Traits can help find patterns within complex communities by canopy traps, both had unexpectedly higher richness and composition grouping species based on characteristics that influence their fitness or relative to the understory, likely suggesting free movement in vertical success (McGill et al. 2006), and may provide insight into the variable space instead of canopy fidelity (Ulyshen et al. 2020). An Ohio, USA stratification patterns we describe above. For example, prior work in study instead found significantlymore bees in the understory than in the similar systems found that bees in understory forests habitats were more subcanopy or canopy, but the two elevated heights had similar com­ social and wood-nesting than in forest openings (Roberts et al. 2017), munity composition (Cunningham-Minnick and Crist 2020). In one and that bees with different diet breadths may be differently limited by observational study of the leguminous tree Adesmia tristis, the authors only forage or also by nesting resources (Westerfelt et al. 2018). In our hypothesized that the Megachile foraging in the upper canopy had work, we focus on bee sociality, nesting habit, and diet breadth, all competitively excluded , which were instead found only in the factors known to influence bee behavior and environmental responses lower canopy (Ferreira et al. 2014). Finally, in one of the only studies of (e.g. Williams et al., 2010; Harrison et al., 2018). vertically stratified nesting, trap-nests installed in 12 German forest Our research was designed to fill several knowledge gaps regarding stands across a gradient of tree diversity found higher abundance in the temperate forest stratification, as well as associations with possible canopy and at sites of higher tree diversity (Sobek et al. 2009). Due to drivers such as local light differences via canopy cover and landscape- variable patterns regarding stratification in temperate forests, further scale nesting resources. We ask: (1) What is the composition of canopy mechanistic work is needed to understand drivers of community pat­ bee communities in the early spring woodlots and forest of NY, USA? (2) terns across strata. Are there patterns of canopy and understory stratification?And (3) does Consistent definitions of the canopy and its boundaries remain bee sociality, nesting habit, or diet breadth govern a bee’s presence in elusive, despite researchers’ intuition of their conceptual existence, and vertical space? We hypothesize that male, brood parasitic, and ground- definitionsoften vary among studies (Moffett, 2001; Basset et al., 2003). nesting bees will be more abundant in the understory. A bee whose diet In the studies discussed above, canopy is often not explicitly defined,but is specialized on certain spring ephemeral flowers,e.g. Andrena erigeniae generally corresponds with the highest level of vegetation, or topmost (Parker et al. 2016) may similarly be more often found in the understory. branches of the trees. These vegetative distinctions between strata may In contrast, bees foraging in the early spring canopy may be seeking non- or may not be important to highly mobile flyinginsects such as bees and floral plant resources (Requier and Leonhardt 2020) or have generalist wasps that are expected to move without constraint between strata. diets (Ramalho 2004). We hypothesize generalized bees may be more Instead, a bee may prefer a given canopy stratum for only a specificlife likely to visit the many tree species in our system that are both wind- stage, time of day, season, resource, or host plant bloom period (Basset pollinated and mast-blooming, so not reliably found each year; while et al. 2003; “opportunistic usage,” sensu Roubik 1993). For example, some wind-pollinated trees may be good protein sources (Roulston and Roubik (1993) found that 20 tropical bee species were not strictly Cane 2000), they often lack nectar as well as a shared evolutionary associated with forest strata, except two nocturnal species found history with wild bees. Finally we explore whether two aspects of consistently in the high-canopy, while a study of stingless bees in Brazil habitat – the volume of woody nesting material or canopy cover at a tree found that canopy bees were correlated with trees’ floweringabundance – affects captured bee abundance. We hypothesize that wood-nesting (Ramalho 2004). These two studies defined canopy vastly differently: bee abundance will be higher in forests with a greater volume of 20–25 m and >7 m, respectively (Roubik, 1993; Ramalho, 2004). coarse woody debris. In the understory, we hypothesize that we will Despite the variable usage of the term canopy, we emphasize the need to catch more bees in plots with less canopy cover and thus more light study higher-elevation habitats and the preferences that guide their use (Williams and Winfree 2013). Finally, we interpret the observed pat­ by organisms of high conservation concern. terns and discuss implications for management. Understanding canopy bees could have important implications for management, as canopy flower diversity and nesting sites would help 2. Methods maintain resource continuity for pollinators prior to crop bloom (Ulyshen, 2011; Iuliano and Gratton, 2020). Yet, most studies of the 2.1. Study system & field methods communities in temperate forests have focused on ground-level fauna, and largely describe small-bodied, solitary, and specialist spe­ Eleven second-growth forest patches were selected across the Finger cies visiting spring ephemerals (Harrison et al., 2018; Smith et al., Lakes region in New York State, USA (Fig. 1). This region is a hetero­ 2019), rather than including vertical gradients (Hanula et al. 2016). geneous matrix of forest, agriculture, and developed land use types Bees sampled only at ground level are often studied in timber harvests (Park et al. 2015). Each woodlot or forest was selected for immediate (McNeil et al. 2019) or forest openings (Roberts et al. 2017), on pow­ adjacency to an apple orchard (or in two cases separated by <300 m by a erline cuts or roads (Russell et al. 2018), or across successional gradients small field), as a longer-term motivation of our work is to explore if (Milam et al., 2018; Odanaka and Rehan, 2020). Most studies find the forest canopy may provide spillover of pollinators to orchards and/or highest abundance of bees in openings, gaps, or edges, presumably biodiversity refuges (Urban-Mead, unpublished data). Sites dominated by responding to micro-site characteristics such as light levels (Williams beech or hemlock stands were excluded as they represent unique eco­ and Winfree 2013). This prior work on ground-level bees has led to systems and were not the focus of this study. Woodlot/forest choice was understory-focused management recommendations, including a focus otherwise representative of regional woodlot and forest patch compo­ on light-filled gaps and openings for pollinator conservation, and sition and management. We sampled the same six forests in 2017, 2018, necessarily little emphasis on the possibility of vertically distributed and 2019. In 2018 we sampled five additional sites, and in 2019 kept offerings (Hanula et al. 2016). Yet, canopy and vertical structural ele­ three of those sites. The total number of sites sampled per year was: ments have been studied at length for other taxa (e.g. Shields et al., 2017: 6 sites; 2018: 11 sites; 2019: 9 sites. Dominant and co-dominant 2008; Verschuyl et al., 2008; Dolek et al., 2009; Hyvarinen¨ et al., 2009; canopy species at all sites were Acer saccharum, Acer rubrum, Quercus Di Giovanni et al., 2015; Hanle et al., 2020) and then incorporated into rubra, Fagus sylvatica, and Betula spp. Subcanopy trees at many sites habitat-oriented management plans (Loehle et al., 2002; Ishii et al., included Ostrya virginiana and Carpinus sp. Several sites also had Tilia 2003). In Japan a correlation between canopy flowering abundance in americana, Populus deltoides, Juglans nigra, Carya ovata, and Carya one year with bumblebee queens the next (Inari et al. 2012) suggests cordiformis.

2 K.R. Urban-Mead et al. Forest Ecology and Management 482 (2021) 118903

Fig. 1. Study design and experimental set-up. Field work was conducted at six sites in 2017, fiveadditional sites in 2018 (total = 11 sites), and three of those sites were kept in 2019 (total = 9 sites). Five trees at each forest site were set up with canopy and understory bee bowl traps on a pulley system and emptied and re-set every 7–10 days.

We used identical sampling effort for bees in the canopy and un­ unique identifiers and associated metadata (Colwell 1994). Specimens derstory. In our study, we define the canopy as the uppermost layer of have been accessioned to the Cornell University Insect Collection. vegetation, 20–25 m high. Five paired canopy-understory trap sets were Specimens were identifiedto species or nearest possible morphospecies installed at ground level (no more than 1 m) at each site below their using online guides and existing taxonomic keys (Laberge, 1971; respective canopy trap. For each trap, a canopy-dominant tree was LaBerge, 1973; 1980; 1985;; Gibbs, 2011; Gibbs et al., 2013, discov­ selected and a BigShot® sling shot was used to rig a canopy trap; tree erlife.org, and Andrena subgeneric keys by Mike Arduser, unpublished). and branch selection were constrained by sling shot access and clear Due to unresolved and lack of available keys, specimens of passage in the vertical space below for raising and lowering the traps. Nomada (n = 61), Hylaeus (n = 3), and Sphecodes (n = 3) were only “Bee bowl” pan traps were custom-built from three spray-painted Solo® determined to the genus level; some individuals were determined to the cups tied vertically with paracord (Fig. 1, right panel); (methods sensu generic or subgeneric level due to damage in bee bowls (n = 35). Other Ulyshen et al. 2010, Cunningham-Minnick and Crist 2020). We used difficult taxa were resolved in two different ways: with the help of sixteen-ounce cups to minimize spills while traps were vertically sus­ taxonomic experts (Jason Gibbs assisted with Lasioglossum (Dialictus)), pended in windy canopies. Bowl colors (yellow, blue, and white) were or using DNA barcoding for a subset of challenging Andrena (see below). randomized vertically, and bowls were filled with soapy water and We extracted DNA from 130 Andrena bees using a CTAB protocol deployed in canopy-emergent trees in all woodlots within 100 m of the available on the Danforth lab website (http://www.danforthlab.entomo forest edge nearest the adjacent orchard. Canopy traps were not visible logy.cornell.edu/research/resources/). These bees were selected as a from the ground. Bowls were emptied and re-set every 7–10 days subset of the full dataset either to confirm a tentative morphological beginning in late March and ending in mid-June, following canopy identification,or because the specimen was damaged or exhibited traits closure. Trees with traps were 100–200 feet apart within the woodlot or that could not be categorized clearly with the existing keys. DNA was forest plot; due to this small spatial scale, bee community data were extracted from 3 legs which allowed us to retain the remaining, nearly pooled by site for most analyses. Bowl traps were poured through complete, specimen for further study and as a pinned voucher. We used strainers in the field, and bees were transferred directly into Whirlpaks primers developed by Creedy et al. (2020) for barcoding British bees: ′ ′ ′ (Nasco Whirlpak, Fort Atkinson, WI) containing 95% ethanol where BeeF (5 -TWY TCW ACW AAY CAT AAA GAT ATT GG-3 ) and BeeR (5 - ′ they were stored until processing in the lab. In 2018 and 2019, canopy TAW ACT TCW GGR TGW CCA AAA AAT CA-3 ). These bee-specific cover was measured at each collection event using a Densiometer from primers worked consistently across nearly all of our samples and ◦ the base of each tree by averaging cover taken facing each of four car­ rarely yielded non-target sequences. For PCR, we used 35 cycles of 94 C ◦ ◦ dinal directions and converted to overstory density percentage. Coarse for 45 s, 52 C for 45 s, and 72 C for 1 min. All PCR fragments were woody debris (CWD) was quantified at the site level along three 50 m sequenced in both directions. After trimming primers, our fragments spoke transects extending from the center point of the fivesampled trees were 658 bp in length. Sequences were edited and aligned in Geneious at each site. We measured length and diameter of both ends of all CWD (Geneious Prime® 2019.0.4; Build 2018-11-27 02:05; Java Version 11 with a diameter greater than 7.62 cm (3in) at the transect intersect; all + 28 (64 bit), available at https://www.geneious.com). CWD encountered was lying on the forest flooror reaching no more than 3 m above ground if lying at an angle. Total volume in cubic feet per site 2.3. Data analysis – Community level was calculated using Smalian’s formula, which assumes a cone-shaped log (Woodall and Monleon 2007). All data analyses were conducted in R version 4.0.1 (2020-06-06) – “See Things Now” (R Core Team 2020). We used Chao’s Rpackage iNEXT 2.2. Laboratory methods to calculate Hill numbers, the “effective number” of species (Chao et al. 2014). This framework standardizes comparisons of different samples by All specimens were pinned and databased using Biota software with implementing rarefaction and then extrapolation, inferring the

3 K.R. Urban-Mead et al. Forest Ecology and Management 482 (2021) 118903 asymptotic number of species under complete sampling (Chao et al. in order to meet assumptions of linear models. A Guassian rather than 2014). Simply, based on a single equation, the exponential value of q Poisson model was appropriate due to a normal distribution (also provides different weights to rare species, where q = 0 weights all because all values were season-long sums, thus larger numbers with no species regardless of abundance and is termed Species Richness, q = 1 zeroes in the dataset). weights species relative to their abundances and generates the index Canopy cover increased with time (R2 = 0.58, p < 0.001), which was “Hill-Shannon,” the exponent of Shannon Diversity, while q = 2 deem­ expected as the leaves grew. To investigate whether canopy cover at a phasizes rare species, and is the inverse of the familiar Simpson Index given sampling event (each tree with a trap on a given day) influenced (Jost 2006, Chao et al. 2014). We drew season-long pooled diversity the number of bees of each sex we caught, we used a Poisson model as accumulation curves, including the extrapolated numbers of bee species these were count data. We ran two separate GLMMs for the canopy and for each value of q in Rpackage ggplot2 using iNEXT’s command ggi­ the understory because canopy cover was only measured in the under­ NEXT() (Chao et al. 2014). After calculating the diversities of our canopy story so we hypothesized that the strata may be affected differently. Sex and understory communities using the Hill numbers, we used general­ and canopy cover were included as fixedeffects and the site as a random ized linear mixed effect models (GLMMs) to ask if bee communities effect, with tree nested within site. differed between strata (canopy or understory), and whether differences To test the hypothesis that bees of different sexes and trait groups were consistent among years. In each model, the fixedeffects were strata may respond differently to the amount of CWD at a site, we used the and year and their interaction, with site as the random effect. We ran command lme() in the Rpackage nlme with the interaction of trait group models with the same structure for bee abundance. by bee abundance as the response and site as a random effect. Canopy To ask if male or female bees were differentially abundant in and understory models were run separately. The response variable was different strata, we used a binomial GLMM (log link) with strata as the log (n + 1) transformed due to heterogeneity of variance among trait predictor, proportion male or female bees as the response variable, and groups. We used the buffer tool to calculate the amount of forest in 750 site as the random effect. Post hoc analyses and log odds ratios were m using the 2018 National Agricultural Statistics Service Cropland Data calculated in Rpackage emmeans. Layer (USDA, National Agricultural Statistics Service 2018) for New We used non-metric multidimensional scaling (NMDS) in Rpackage York in qGIS (QGIS Development Team 2020). Buffers were drawn vegan with command metaMDS() to visualize the difference in bee around the centroid of our five sampled trees of each woodlot as this communities in the canopy and the understory. This analysis used the distance is known to influence the smaller bodied and solitary bee full dataset of all species, but with male and female bees considered populations that dominated our community (Steffan-Dewenter et al. uniquely. We applied the Hellinger transformation to the matrix, which 2002). We used models with CWD multiplied by the proportion of the takes the square root of row totals, thus reducing the weight of rare landscape that was forested in the buffer around each site (termed species and super-abundant species. This transformation is appropriate “landscape-scaled CWD” hereafter), as a proxy for the relative amount of for skewed ecological data (Legendre and Gallagher 2001) such as ours CWD available at the landscape scale. Finally, we ran these models again (Fig. A.2). We confirmedhomogeneity of variances using betadisper() in with only the CWD site-level values and only the proportion of forest, to the Rpackage vegan before running the adonis tests with sex and strata make sure that any trends found by the scaled CWD metric were not (canopy or understory) as group levels (p > 0.1 for all data trans­ simply driven by a strong forest-cover trend. formations, thus failing to reject the null hypothesis of homogeneous variances). Then, we performed a PERMANOVA with 999 permutations 3. Results in the Rpackage pairwiseAdonis (Martinez Arbizu 2020), with the sex and strata (canopy or understory) as fixedeffects and site as a random effect 3.1. Results – Community level using the argument strata. Pairwise adonis automatically implements a Bonferroni correction for multiple comparisons with p.adjust(). Finally, We collected a total of 2833 bees over three years, of which 554 were we conducted an indicator species analysis using command multipatt() sampled at 6 sites in 2017, 1171 at 11 sites in 2018, and 1108 at 9 sites in in Rpackage indicspecies to ask if specific bee species were associated 2019. Of these bees, 1365 were caught in canopy traps, and 1468 bees with either stratum. We ran this analysis once with the full community were caught in the understory. In the canopy, 79.2% of the bees were and again with rare bees excluded to ensure they were not biasing the female, while in the understory female bees comprised 59.0%. Female results (definedin this case as species-sex combinations with singletons bees were significantlymore associated with the canopy than male bees and doubletons at both strata). (Fig. 2.), with the log odds ratio of females 2.66 times that of males (SE = 0.24, p < 0.0001). Across all sites, we found 90 species out of 417 bee 2.4. Data analysis – Trait and site predictors of bee communities species known in New York (Danforth and van Dyke 2015) (DNA bar­ coding results A.1; species lists, Appendix B). Of these, 82 species were We categorized bees by trait groupings of nesting habit, sociality, found in the canopy and 81 in the understory. The five most common and diet breadth using pre-existing databases (Bartomeus et al., 2013; bees overall were male and female Andrena rugosa (n = 369 and 124, Normandin et al., 2017) and expert knowledge (B.N.D. and K.R.U.). respectively), female Lasioglossum coeruleum (n = 339), female Augo­ Traits such as these are likely to influencehow an organism responds to chlora pura (n = 221), female L. versans (n = 151) and female its environment (McGill et al. 2006). Diet breadth was categorized as L. quebecense (n = 98). All but 20 individual bees of 5 bee species (all in generalized or specialized (polylectic or oligolectic following Bartomeus the genus Andrena) were classifiedas polylectic, or diet generalists, and et al. 2013). Sociality was categorized as social, solitary, or facultatively so this trait was not included in further analyses. social, with brood parasites excluded (due to their contrasting life his­ Bee species abundances were highly right-skewed, which is common tory Danforth et al. 2019). Nesting habits were cavity, soil, wood, hole, for ecological community data; species accumulation curves showed and stem (Bartomeus et al. 2013), and cleptoparasitic. Following these adequate sampling (Fig. A.2). Overall, we found that bee abundance did literature categorizations, cavity refers to pre-existing cavities typically not differ significantlybetween the canopy and understory (strata: F1,36 used by bumble bees and honey bees, while hole refers to cavity usage = 0.43, p = 0.52, year: F2,36 = 1.59, p = 0.22), but that there was a primarily by Megachilid bees. significantinteraction between year and strata (F2,36 = 4.33, p = 0.02); We used linear models to assess whether the abundance of bees in bees were less abundant in the canopy in 2017 and 2019, but more each trait group differed by strata (canopy vs. understory). Models were abundant in 2018 (Fig. A.3). There was also no difference in Species run separately by bee sex, with trait group (nesting habit or sociality), Richness between strata or years, nor any interaction with year (strata: strata and their three-way interaction as fixed effects, and site nested F1,36 = 1.15, p = 0.29, year: F2,36 = 2.21, p = 0.12, interaction: F2,36 = within year as the random effect. Response variable was log-transformed 1.20, p = 0.31) (Fig. A.3). However, Hill-Shannon diversity was higher

4 K.R. Urban-Mead et al. Forest Ecology and Management 482 (2021) 118903

Fig. 2. Female and male bees differ in abundance in canopy strata. Left: Total number of individual bees per site over all three years. The boxplots show the median, first and third quartiles, and the points are outliers; significance determined by Tukey two-sided post-hoc tests. Right: Annual variation in mean number of male and fe­ male bees caught per site in each forest strata. 2017, n = 6 sites; 2018, n = 11 sites; 2019, n = 9 sites. In both panels, light gray are female bees, dark gray are male bees.

in the canopy (strata: F1,36 = 4.70, p = 0.03) and differed among years There was a significant three-way interaction in both models (sociality (year: F2,36 = 5.00, p = 0.01), but the greater diversity in the canopy was F2,352,=7.01, p = 0.001; nest habitat F5,736,=3.64, p = 0.0029). We note consistent across years (interaction: F2,36 = 1.72, p = 0.19) (Fig. A.3). that there were no male social or wood-nesting bees (Fig. 4). Ordination visually corroborated that community composition differed In all linear models asking if trait groups responded differently to between strata and sexes (PERMANOVA, F1,43 = 3.15, p = 0.02 site-level CWD, abundance was different among trait groups (p < 0.001 and F1,43 = 24.13, p = 0.001, respectively; there was no interaction ef­ in all models), in agreement with the prior analysis (Table A.4). We fect) (Fig. 3). Results were qualitatively similar when run on presence- found no main effect of forest-scaled CWD on overall canopy or under­ absence data, and similar with raw abundance data although with an story abundance (canopy F1,9 = 0.26, p = 0.63; understory F1,9 = 0.81, p additional significant interaction between sex and strata (all output in = 0.39), also consistent with the above. Canopy abundance was not Table A.4). correlated with forest-scaled CWD nor were any interactions significant Indicator species analysis by sex and taxa did not identify any of the (p > 0.1 for all). In the understory, bees with different nesting habits most abundant bees as strata-affiliates. Canopy indicator species were responded differently to forest-scaled CWD (significant interaction; Andrena perplexa males (15 out of 17 individuals in canopy), A. nivalis F4,36 = 4.31, p = 0.006). In the model with bees divided by sociality, females (7 out of 8 in canopy), and Osmia pumila females (11 out of 11 in increasing forest-scaled CWD showed a positive trend with understory canopy), while the only understory indicator was A. tridens males (33 abundance (F1,9 = 4.72, p = 0.058), but responses did not differ between out of 47 in understory; all p < 0.05 with 999 permutations; see dis­ trait groups (no significant interaction; F2,18 = 2.73, p = 0.09). We ran cussion of species determination of A. tridens in Fig. A.1). Results were these same tests with CWD and forest proportion each alone as the the same when run with rare species excluded. predictor variables to ask if either of these were driving the scaled trend. In the understory, there was a significantinteraction of nesting habit and 3.2. Results – Trait and site predictors of bee communities forest proportion (F4,36 = 2.62, p = 0.0503), although it was a weaker relationship than when the predictor was scaled by CWD. We found no In both trait-based linear models, abundance of bees was signifi­ other effects of either predictor on overall abundance, sociality, or nest cantly different by sex and trait (sociality: sex F1,352,= 148.43, p < habit (p > 0.1 for all tests). 0.0001, trait F2,352,=211.16, p < 0.0001; nest habitat: sex F5,736,=70.17, More canopy cover was correlated with fewer overall bees caught at p < 0.0001 , trait F5,736,=178.50, p < 0.0001), while abundance did not a given tree and sampling location in the understory (z = -16.46, p < differ between strata, consistent with the above abundance models 0.0001), but not in the canopy (z = 0.15, p = 0.88) (Fig. 4). Male bees (sociality F1,352,=0.61, p = 0.44, nest habitat F1,736,=2.87, p = 0.09). In were more negatively correlated to canopy closure than female bees in both models, the abundance of male and female bees of different traits both strata (interaction term for canopy z = -6.32, p < 0.0001; under­ were significantlydifferent between strata (sociality: sex*canopy strata story z = -5.43, p < 0.000). When tested separately, increased canopy F1,352,= 11.15, p < 0.001, trait*canopy strata F2,352,=211.16, p < cover was correlated with fewer female and male bees in the understory 0.0001; nest habitat: sex*canopy strata F1,736,=10.33, p = 0.001, (female z = -8.58, p < 0.0001, male z = -13.97, p < 0.0001), and an trait*canopy strata F5,736,=2.58, p = 0.025). Finally, bees of different increase in canopy females but a decrease in canopy males (female z = sexes responded differently depending on their trait groups (sociality 4.17, p < 0.0001, male z = -6.08, p < 0.0001; Fig. 4). F1,352,=43.24, p < 0.0001, nest habitat F5,736,=48.29, p < 0.0001).

5 K.R. Urban-Mead et al. Forest Ecology and Management 482 (2021) 118903

Fig. 3. NMDS plot of bees separated by sex and strata. Data were Hellinger transformed before being plotted using Bray-Curtis distances (ordination solution stress = 0.157). Points are fieldsites; ellipses are at 95% confidenceinterval, and are drawn separately for each stratum and sex. Left panel is all years. Right stacked panels: each year presented separately as labeled.

4. Discussion The most abundant taxa in our study had strata preferences during our study period, although in contrast to previous studies, the canopy We found an abundant, diverse assemblage of predominantly female communities we studied were not dominated by one or two single spe­ bees in the springtime canopy of the northeastern temperate hardwood cies (Ulyshen et al., 2010; Campbell et al., 2018). It may be that if we forest woodlots. Indeed, over 20% of New York’s 417 bee species were sampled later in the summer, as in Ulyshen et al. 2010, that dominance found active in the forest canopy. We found similar species richness at would have emerged. Particularly, 73% of Lasioglossum (Dialictus) both heights, and no evidence of any bee species exclusively bound to a coeruleum and 62% of Augochlora pura were in the canopy. Both are specific forest stratum (understory ~ 1 m, or canopy 20–25 m). Abun­ iridescent halictids known to nest in rotting logs (Stockhammer, 1966; dance varied year to year, consistent with other studies finding high 1967), and it has been suggested that A. pura may be collecting hon­ intra-year community-level variation in bee diversity and abundance (e. eydew (Ulyshen 2011). Two other abundant bees, Andrena rugosa and g. Alarcon´ et al., 2008; Dupont et al., 2009), which in our system may be L. (Dialictus) versans, were disproportionately in the understory. Lasio­ particularly driven by year to year weather or resource variability due to glossum versans is a previously identified forest associate (Giles and mast bloom (Inari et al. 2012). However, female wood-nesting and fe­ Ascher, 2006; Romey et al., 2007; Roberts et al., 2017), while to our male social bees were consistently more abundant in the canopy than in knowledge A. rugosa has not been previously identified as a forest- the understory and as we hypothesized, the understory was instead associated bee. The high proportion of A. rugosa males in the under­ dominated by males of solitary, soil-nesting species. Diet breadth did not story (~84%) suggests a soil-nesting habit consistent with all other differ between strata, which were both dominated by generalist bees. known Andrena; future research should explore whether it prefers spe­ Female bees’ primary activity is foraging, which suggests more attention cificmesic forest soils such as in our system. Finally, nearly 80% of the should be paid to canopy pollen, nectar (e.g. Chambers 1946, Batra cleptoparasitic genus Nomada were caught in the understory, corrobo­ 1985, reviewed in Saunders, 2018; e.g. abundant maple pollen collec­ rating our hypothesis that brood parasites of ground-nesting bees would tion by Colletes inaequalis, Nicholas Dorian personal communication) or be abundant on the forest floor. even honeydew foraging (Trelease, 1881; Requier and Leonhardt, Coarse woody debris (CWD) and canopy closure were only partially 2020). If canopy resources are thus widely used, we recommend man­ associated with bee community metrics. Other studies have also found agement for diversity of tree species, bloom time, and age to ensure limited correlations at the site level (Sobek et al., 2009; Simanonok and spatial and temporal continuity in the availability of canopy pollen re­ Burkle, 2019). One possible interpretation for the lack of relationship is sources for bees, flower-flies, and other pollen-foragers (Iuliano and that our study sites may have sufficient deadwood for nesting. Alter­ Gratton 2020), and retention of deadwood nesting habitat (Freedman natively, if nesting in canopy branches or standing snags is common, an et al., 1996; Westerfelt et al., 2018). above-ground quantification of deadwood would have been more

6 K.R. Urban-Mead et al. Forest Ecology and Management 482 (2021) 118903

Fig. 4. Bee abundance by canopy cover and trait. Left panel: Number of bees caught at a given tree relative to the proportion canopy cover as measured at ground level with a densiometer. Counts of female bees light gray, male bees dark gray; lines drawn from Poisson GLMM, with shading 95% confidence interval. Middle & Right panels: Abundance of bees by nesting habit and degree of sociality in canopy and understory. Boxplots show the median, firstand third quartiles, and the points are outliers. Letters are applied independently for contrasts within the trait categories. Light gray boxes are females, dark gray boxes are males; there were no male social or wood-nesting bees in our dataset due to phenology. Groups within each trait category and strata that do not share a letter are significantlydifferent; letters generated with compact letter display. relevant than understory CWD for explaining canopy abundance. This abundance. could similarly explain why canopy bee abundance did not decline as To better understand the bee community despite these limitations the canopy closed (also hypothesized in Ulyshen et al. 2010). Another (Portman et al. 2020), we also attempted multiple days of canopy Ohio, USA study also found many wood-nesting bees in the canopy — netting on open flowers, and observed phenomena worthy of future these authors even suggest die-offs due to Emerald Ash Borer, Agrilus exploration. Large bees, including Xylocopa and Bombus, were planipenis, could have increased wood-nesting bee habitat (Cunning­ frequently observed foraging during canopy climbs, but qualitatively ham-Minnick and Crist 2020). We also note that late-spring canopy were much more common on the outermost flowersof the canopy, with closure co-occurred with the bloom period of oak trees (Quercus spp.) smaller bees and flowerflies dominant at the interior of each tree, which and later tuliptree (Liriodendron tulipifera) and linden (Tilia americana), may be due to windiness. Bombus spp. queens were also commonly where females might instead have been increasing their foraging and observed in nest-searching behavior in understory leaf litter (e.g. explaining the increase in canopy female bees as the canopy closed. Pugesek and Crone 2020). Bees in the families and Hal­ More work is needed to disentangle these possible explanations. In ictidae dominate bee bowl collections, while Colletidae and Apidae contrast, the observed decline in understory abundance with canopy (including Bombus and Apis) are under-sampled in bee bowl schemes closure is consistent with prior studies finding higher abundance in (Grundel et al. 2011). Our bee bowl data thus likely underestimate the light-filledmicrosites (Proctor et al., 2012; Williams and Winfree, 2013), bee diversity of apid bees active in both forest strata. In several instances and a general decrease in insect biodiversity with denser mature canopy a single canopy collection produced many Ceratina, possibly suggesting (Hilmers et al. 2018). a local nest site, although this would be quite surprising as Ceratina nest There are multiple critiques of bee bowls, many of which we share. in pithy stems (Rehan and Richards 2010). Second, while canopy tree One concern is that bees may be more attracted to bowls in low- or poor- climbing we observed several vibrant Augochlorini nests in dead oak resource habitats, and ignore them when resources are abundant branches at ~25 m aboveground (K. Urban-Mead, personal observa­ (summarized in Portman et al. 2020). For example, we sampled very few tion). These startling anecdotal findssupport the above hypothesis that understory floral specialist bees, such as Andrena erigeniae or wood-nested bees may nest much higher above ground than usually A. uvulariae, which may simply be an artifact of those bees not investi­ considered (Stockhammer 1966). gating bowl traps but instead visiting their specialist host flowers It is likely that the early spring wind-dispersed pollen is an important (respectively Claytonia virginiana and Uvularia sessilifolia). Still, we are resource for bees in temperate forests, and efforts should be made to confidentthat canopy bowls were not at all visible from the forest floor, maintain diverse plant communities for bees — not only understory and so infer that bees must have already been in the canopy and were flowering resources but also early spring blooming trees. In the early not visually attracted from the understory. In parallel sampling (un­ spring, male bees emerge earlier in solitary species while female bees published), bowls placed in apple orchards caught no bees until the crops proliferate earlier in social species (Danforth et al. 2019). Andrena rugosa began to bloom, in that case faithfully representing relative local was among the solitary species for whom we caught both sexes of

7 K.R. Urban-Mead et al. Forest Ecology and Management 482 (2021) 118903 important common crop pollinators (Renauld et al., 2016; Blitzer et al., Lasioglossum: Dialictus identification; Marlyse DuGuid and D. J. McNeil 2016; Grab et al., 2019), as well as bees in the subgenus Melandrena for advice on coarse woody debris sampling; and Erika Mudrak at the (specificallyA. carlini, A. pruni, and A. dunningi) which has been shown Cornell Statistical Consulting Unit for statistical support. This work was to have extremely high per-visit pollination efficiencyfor apple flowers supported by the National Science Foundation Graduate Research (Park et al., 2016; Blitzer et al., 2016). Our data suggest that the male life Fellowship under Grant No. 1650441 and SARE Graduate Student Grant stage of many of these important crop pollinators, which are not usually GNE18-188. Any opinions, findings, and conclusions or recommenda­ emphasized habitat evaluations, may be particularly abundant in for­ tions expressed in this material are those of the authors(s) and do not ests. We recommend future research should identify overlapping pollen necessarily reflect the views of the National Science Foundation. I usage and preference of bees in adjacent agricultural habitats, as forest particularly would like to thank all of the growers and land managers management for bees may protect these wild pollination services who welcomed me onto their farms and from whom I learn so much. (Winfree et al. 2007). Finally, I acknowledge that my fieldsites, home, and Cornell University Our characterization of high diversity and vibrant canopy bee ac­ are on unceded land of the Gayogoho:´ n people (Cayuga Nation) of the tivity in early spring recommends inclusion of wild bees’ vertical Haudenosaunee Confederacy. behavior in forest management plans (Ulyshen, 2011). Beyond impor­ tant recommendations for light-filled gaps and habitat heterogeneity Appendix A. Supplementary data (Hanula et al. 2016), our results additionally emphasize the high di­ versity of female wood-nesting bees caught in the canopy. We reiterate Supplementary data to this article can be found online at https://doi. recommendations for practicing retention forestry and keeping high org/10.1016/j.foreco.2020.118903. stumps and snags as habitat (Westerfelt et al. 2018). Generally, more species-diverse forests support more diversity across taxa (Ampoorter References et al. 2020), and previous work has demonstrated that canopy bloom ´ abundance can have species-specific demographic implications (Inari Alarcon, R., Waser, N.M., Ollerton, J., 2008. Year-to-year variation in the topology of a plant-pollinator interaction network. Oikos 117 (12), 1796–1807. et al. 2012). Honeydew collecting bee species will benefitfrom presence Ampoorter, E., Barbaro, L., Jactel, H., Baeten, L., Boberg, J., Carnol, M., Castagneyrol, B., of honeydew trees such as Populus, Betula, and Quercus, and possibly Charbonnier, Y., Dawud, S.M., Deconchat, M., Smedt, P.D., Wandeler, H.D., ¨ even Tilia, Acer, and Platanus (Crane and Walker 1985), and future work Guyot, V., Hattenschwiler, S., Joly, F.-X., Koricheva, J., Milligan, H., Muys, B., Nguyen, D., Ratcliffe, S., Raulund-Rasmussen, K., Scherer-Lorenzen, M., van der should continue to explore the anecdotal records of anemophilous and Plas, F., Keer, J.V., Verheyen, K., Vesterdal, L., Allan, E., 2020. Tree diversity is key ambophilous pollen collection (Saunders, 2018). We suggest that man­ for promoting the diversity and abundance of forest-associated taxa in Europe. Oikos agement for tree species and age class diversity will maintain continuity 129 (2), 133–146. of pollen and honeydew resource availability from diverse tree species. Bartomeus, I., Ascher, J.S., Gibbs, J., Danforth, B.N., Wagner, D.L., Hedtke, S.M., Winfree, R., 2013. Historical changes in northeastern US bee pollinators related to This is particularly important as tree species are often only in bloom for shared ecological traits. Proc. Natl. Acad. Sci. 110 (12), 4656–4660. very short periods, and not consistently each year (e.g. Houle, 1999; Basset, Y., Kitching, R., Miller, S., Novotny, V., 2003. Arthropods of Tropical Forests: Fernandez-Martínez´ et al., 2012). Thus, as appropriate, varied silvicul­ Spatio-Temporal Dynamics and Resource Use in the Canopy. Cambridge University Press. tural treatments aimed at maintaining different age stands and Batra, S.W.T., 1985. Red maple (Acer rubrum L.), an important early spring food resource mimicking natural disturbances should be considered to maintain age for honey bees and other insects. J. Kansas Entomol. Soc. 58, 169–172. class and species diversity at both the patch and landscape scale. These Blitzer, E.J., Gibbs, J., Park, M.G., Danforth, B.N., 2016. Pollination services for apple are dependent on diverse wild bee communities. Agric. Ecosyst. Environ. 221, 1–7. recommendations are consistent with strategies to maintain avian di­ Campbell, J. W., P. A. Vigueira, C. C. Viguiera, and C. H. Greenberg. 2018. The effects of versity (e.g. Duguid et al. 2016, Hanle et al. 2020), which often partic­ repeated prescribed fireand thinning on bees, wasps, and other flowervisitors in the ularly emphasize vertical structure (e.g. King and DeGraaf 2000, understory and midstory of a temperate forest in North Carolina. Forest Science 64: 299–306. Goodale et al. 2009). The multiple niches generated for nesting and Chambers, V.H., 1946. An examination of the pollen loads of andrena: The species that forage may then support social species in the process establishing their visit fruit trees. J. Anim. Ecol. 15 (1), 9. https://doi.org/10.2307/1621. nests, important solitary crop pollinating-bees such as Andrena, and Chao, A., Gotelli, N.J., Hsieh, T.C., Sander, E.L., Ma, K.H., Colwell, R.K., Ellison, A.M., 2014. Rarefaction and extrapolation with Hill numbers: a framework for sampling nesting habitat for iridescent pollinators including the CWD- and snag- and estimation in species diversity studies. Ecol. Monogr. 84 (1), 45–67. nesting green and blue sweat bees. Finally, we recommend future Chapin, K.J., Smith, K.H., 2019. Vertically stratified diversity in a Florida work on bee canopy communities in temperate hardwood forest in­ upland hardwood forest. Florida Entomologist 102, 211–215. teriors, which may have different patterns of vertical stratificationthan Colwell, R.K., 1994. Biota : the biodiversity database manager. http://viceroy.eeb.uconn. edu/Biota/. our smaller orchard-adjacent plots (e.g. a study in rainforests found Crane, E., Walker, P., 1985. Important honeydew sources and their honeys. Bee World 66 stronger vertical stratificationof light, temperature, and vapor pressure (3), 105–112. deficit farther from the forest edge (Didham and Ewers 2014)). We Creedy, T.J., Norman, H., Tang, C.Q., Qing Chin, K., Andujar, C., Arribas, P., O’Connor, R.S., Carvell, C., Notton, D.G., Vogler, A.P., 2020. A validated workflow enthusiastically encourage the ecological management of forests to for rapid taxonomic assignment and monitoring of a national fauna of bees explicitly include wild pollinator communities. (Apiformes) using high throughput DNA barcoding. Mol. Ecol. Resour. 20 (1), 40–53. Cunningham-Minnick, M.J., Crist, T.O., 2020. Floral resources of an invasive shrub alter Declaration of Competing Interest native bee communities at different vertical strata in forest-edge habitat. Biol. Invasions 22 (7), 2283–2298. The authors declare that they have no known competing financial Danforth, B.N., van Dyke, M., 2015. The wild bees of new york: our insurance policy against honey bee decline. New York State Fruit Quarterly 23, 17–22. interests or personal relationships that could have appeared to influence Danforth, B.N., Minckley, R.L., Neff, J.L., Fawcett, F., 2019. The Solitary Bees: Biology, the work reported in this paper. Evolution. Princeton University Press, Conservation. Dejean, A., Corbara, B., 2003. A review of mosaics of dominant ants in rainforests and plantations. Page in Y. Basset, V. Novotny, S. E. Miller, and R. L. Kitching, editors. Acknowledgements Arthropods of Tropical Forests: Spatio-temporal Dynamics and Resource Use in the Canopy. We would like to thank Mike Ulyshe, Jessica Forrest, and two Didham, R.K., Ewers, R.M., 2014. Edge effects disrupt vertical stratification of microclimate in a temperate forest canopy. Pac. Sci. 68 (4), 493–508. anonymous reviewers for comments that improved an earlier version of Dolek, M., Freese-Hager, A., Bussler, H., Floren, A., Liegl, A., Schmidl, J., 2009. Ants on this manuscript; Erin Krichilsky, Xavier Carroll, Edward Walter, Andrea oaks: effects of forest structure on species composition. J. Insect. Conserv. 13 (4), Jaruffe for help processing, sorting, dissecting specimens in the lab; Erin 367–375. Duguid, M.C., Morrell, E.H., Goodale, E., Ashton, M.S., 2016. Changes in breeding bird Krichilsky and Andrea Jaruffe for morphospecies sorting and some abundance and species composition over a 20 year chronosequence following identification; Greg Rothman, Tanvi Naidu, Dora Tan, Weston Forster, shelterwood harvests in oak-hardwood forests. For. Ecol. Manage. 376, 221–230. and Laura Kaminsky for help with fieldwork; Jason Gibbs for help with

8 K.R. Urban-Mead et al. Forest Ecology and Management 482 (2021) 118903

Dupont, Y.L., Padron,´ B., Olesen, J.M., Petanidou, T., 2009. Spatio-temporal variation in Martinez Arbizu, P., 2020. pairwiseAdonis: pairwise multilevel comparison using adonis. the structure of pollination networks. Oikos 118 (8), 1261–1269. R package version 4. Fernandez-Martínez,´ M., Belmonte, J., Maria Espelta, J., 2012. Masting in oaks: Mcgill, B., Enquist, B., Weiher, E., Westoby, M., 2006. Rebuilding community ecology Disentangling the effect of flowering phenology, airborne pollen load and drought. from functional traits. Trends Ecol. Evol. 21 (4), 178–185. Acta Oecologica 43, 51–59. McNeil, D.J., Otto, C.R.V., Moser, E.L., Urban-Mead, K.R., King, D.E., Rodewald, A.D., Ferreira, N.R., Franke, L.B., Blochtein, B., Ferreira, N.R., Franke, L.B., Blochtein, B., Larkin, J.L., 2019. Distance models as a tool for modelling detection probability and 2014. Pollen-ovule relation in Adesmia tristis and reflectionson the seed–ovule ratio density of native bumblebees. J. Appl. Entomol. 143 (3), 225–235. by interaction with pollinators in two vertical strata. Anais da Academia Brasileira Milam, J.C., Litvaitis, J.A., Warren, A., Keirstead, D., King, D.I., 2018. Bee assemblages in de Ciˆencias 86, 1327–1336. managed early-successional habitats in southeastern new hampshire. Northeastern Freedman, B., Zelazny, V., Beaudette, D., Fleming, T., Johnson, G., Flemming, S., Naturalist 25 (3), 437–459. Gerrow, J.S., Forbes, G., Woodley, S., 1996. Biodiversity implications of changes in Moffett, M.W., 2001. The nature and limits of canopy biology. Selbyana 22, 155–179. the quantity of dead organic matter in managed forests. Environ. Rev. 4 (3), Normandin, E.,´ Vereecken, N.J., Buddle, C.M., Fournier, V., 2017. Taxonomic and 238–265. functional trait diversity of wild bees in different urban settings. PeerJ 5:e3051. Gibbs, J. 2011. Revision of the metallic Lasioglossum (Dialictus) of eastern North Odanaka, K.A., Rehan, S.M., 2020. Wild bee distribution near forested landscapes is America (: Halictidae: Halictini). Zootaxa 3073:1. dependent on successional state. For. Ecosyst. 7 (1) https://doi.org/10.1186/ Gibbs, J., Packer, L., Dumesh, S., Danforth, B.N., 2013. Revision and reclassification of s40663-020-00241-4. Lasioglossum (Evylaeus), L. (Hemihalictus) and L. (Sphecodogastra) in eastern North Ozanne, C.M.P., Anhuf, D., Boulter, S.L., Keller, M., Kitching, R.L., Korner,¨ C., Meinzer, F. America (Hymenoptera: Apoidea: Halictidae). Zootaxa 3672, 1–117. C., Mitchell, A.W., Nakashizuka, T., Dias, P.L.S., Stork, N.E., Wright, S.J., Giles, V., Ascher, J.S., 2006. A survey of the bees of the black rock forest preserve, New Yoshimura, M., 2003. Biodiversity meets the atmosphere: a global view of forest York (Hymenoptera: Apoidea). J. Hymenoptera Res 15, 24. canopies. Science 301, 183–186. Di Giovanni, F., Cerretti, P., Mason, F., Minari, E., Marini, L., 2015. Vertical stratification Park, M.G., Blitzer, E.J., Gibbs, J., Losey, J.E., Danforth, B.N., 2015. Negative effects of of ichneumonid wasp communities: the effects of forest structure and life-history pesticides on wild bee communities can be buffered by landscape context. Proc. R. traits: Vertical stratification of ichneumonids. Insect Sci. 22 (5), 688–699. Soc. B. 282 (1809), 20150299. https://doi.org/10.1098/rspb.2015.0299. Goodale, E., Lalbhai, P., Goodale, U.M., Ashton, P.M.S., 2009. The relationship between Park, M.G., Raguso, R.A., Losey, J.E., Danforth, B.N., 2016. Per-visit pollinator shelterwood cuts and crown thinnings and the abundance and distribution of birds in performance and regional importance of wild Bombus and Andrena (Melandrena) a southern New England forest. For. Ecol. Manage. 258 (3), 314–322. compared to the managed honey bee in New York apple orchards. Apidologie 47 (2), Grab, H., Branstetter, M.G., Amon, N., Urban-Mead, K.R., Park, M.G., Gibbs, J., Blitzer, E. 145–160. J., Poveda, K., Loeb, G., Danforth, B.N., 2019. Agriculturally dominated landscapes Parker, A.J., Williams, N.M., Thomson, J.D., 2016. Specialist pollinators deplete pollen in reduce bee phylogenetic diversity and pollination services. Science 363 (6424), the spring ephemeral wildflower Claytonia virginica. Ecol. Evol. 6, 5169–5177. 282–284. Portman, Z.M., Bruninga-Socolar, B., Cariveau, D.P., 2020. The State of Bee Monitoring Grundel, R., Frohnapple, K.J., Jean, R.P., Pavlovic, N.B., 2011. Effectiveness of Bowl in the United States: A Call to Refocus Away From Bowl Traps and Towards More Trapping and Netting for Inventory of a Bee Community. env. entom. 40 (2), Effective Methods. Annals of the Entomological Society of America:saaa010. 374–380. Prinzing, A., and S. Woas. 2003. Habitat use and stratificationof Collembola and oribatid Halaj, J., Ross, D.W., Moldenke, A.R., 1998. Habitat structure and prey availability as mites prinzing. Page Arthropods of Tropical Forests: Spatio-Temporal Dynamics and predictors of the abundance and community organization of spiders in western Resource Use in the Canopy. Cambridge University Press. oregon forest canopies. J. Arachnol. 26, 203–220. Proctor, E., Nol, E., Burke, D., Crins, W.J., 2012. Responses of insect pollinators and Hanle, J., Duguid, M.C., Ashton, M.S., 2020. Legacy forest structure increases bird understory plants to silviculture in northern hardwood forests. Biodivers Conserv 21 diversity and abundance in aging young forests. Ecol. Evol. 10 (3), 1193–1208. (7), 1703–1740. Hanula, J.L., Ulyshen, M.D., Horn, S., 2016. Conserving pollinators in North American Pucci, T. 2008. A comparison of the parasitic wasps (Hymenoptera) at elevated versus forests: a review. Nat. Areas J. 36 (4), 427–439. ground yellow pan traps in a beech-maple forest:116. Harrison, T., Gibbs, J., Winfree, R., 2018. Forest bees are replaced in agricultural and Pugesek, G., Crone, E.E., 2020. Contrasting effects of land cover on nesting habitat use urban landscapes by native species with different phenologies and life-history traits. and reproductive output for bumble bees. bioRxiv:2020.09.26.315044. Glob. Change Biol. 24 (1), 287–296. QGIS Development Team, 2020. QGIS Geographic Information System. Open Source Hilmers, T., Friess, N., Bassler,¨ C., Heurich, M., Brandl, R., Pretzsch, H., Seidl, R., Geospatial Foundation Project. Müller, J., Butt, N., 2018. Biodiversity along temperate forest succession. J. Appl. R Core Team, 2020. R: A language and environment for statistical computing. R Ecol. 55 (6), 2756–2766. Foundation for Statistical Computing, Vienna, Austria. Horvath,´ R., Lengyel, S., Szinetar,´ C., Jakab, L., 2005. The effect of prey availability on Ramalho, M., 2004. Stingless bees and mass flowering trees in the canopy of Atlantic spider assemblages on European black pine (Pinus nigra) bark: spatial patterns and Forest: a tight relationship. Acta Botanica Brasilica 18:37–47. guild structure. Can. J. Zool. 83 (2), 324–335. Rehan, S.M., Richards, M.H., 2010. Nesting biology and subsociality in Ceratina Houle, G., 1999. Mast seeding in Abies balsamea , Acer saccharum and Betula calcarata (Hymenoptera: Apidae). The Canadian Entomologist 142:65–74. alleghaniensis in an old growth, cold temperate forest of north-eastern North Renauld, M., Hutchinson, A., Loeb, G., Poveda, K., Connelly, H., 2016. Landscape America. J. Ecol. 87 (3), 413–422. Simplification Constrains Adult Size in a Native Ground-Nesting Bee. PloS one 11: Hyvarinen,¨ E., Kouki, J., Martikainen, P., 2009. Prescribed fires and retention trees help e0150946. to conserve beetle diversity in managed boreal forests despite their transient Requier, F., Leonhardt, S.D., 2020. Beyond flowers:including non-floralresources in bee negative effects on some beetle groups. Insect Conserv. Diversity 2:93–105. conservation schemes. J. Insect Conserv. 24 (1), 5–16. Inari, N., Hiura, T., Toda, M.J., Kudo, G., Bronstein, J., 2012. Pollination linkage between Roberts, H.P., King, D.I., Milam, J., 2017. Factors affecting bee communities in forest canopy flowering,bumble bee abundance and seed production of understorey plants openings and adjacent mature forest. For. Ecol. Manage. 394, 111–122. in a cool temperate forest. J. Ecol. 100 (6), 1534–1543. Romey, W.L., Ascher, J.S., Powell, D.A., Yanek, M., 2007. Impacts of logging on Ishii, H.T., Tanabe, S., Hiura, T., 2003. Exploring the relationships among canopy midsummer diversity of native bees (Apoidea) in a northern hardwood forest. structure, stand productivity, and biodiversity of temperate forest. Ecosystems 50, J. Kansas Entomol. Soc. 80 (4), 327–338. 342–355. Roubik, D.W., 1993. Tropical pollinators in the canopy and understory: field data and Iuliano, B., Gratton, C., 2020. Temporal Resource (Dis)continuity for Conservation theory for stratum? preferences? J. Insect Behav. 6 (6), 659–673. Biological Control: From Field to Landscape Scales. Frontiers in Sustainable Food Roulston, T.H., Cane, J.H., 2000. Pollen nutritional content and digestibility for . Systems 4. Pages 187–209 Pollen and Pollination. Springer, Vienna. Jost, L., 2006. Entropy and diversity. Oikos 113 (2), 363–375. Russell, K.N., Russell, G.J., Kaplan, K.L., Mian, S., Kornbluth, S., 2018. Increasing the King, D.I., DeGraaf, R.M., 2000. Bird species diversity and nesting success in mature, conservation value of powerline corridors for wild bees through vegetation clearcut and shelterwood forest in northern New Hampshire, USA. For. Ecol. management: an experimental approach. Biodivers Conserv 27 (10), 2541–2565. Manage. 129 (1-3), 227–235. Saunders, M.E., 2018. Insect pollinators collect pollen from wind-pollinated plants: Klimes, P., Fibich, P., Idigel, C., Rimandai, M., 2015. Disentangling the diversity of implications for pollination ecology and sustainable agriculture. Insect. Conserv. arboreal ant communities in tropical forest trees. PLoS ONE 10:e0117853. Divers. 11 (1), 13–31. Laberge, W.E., 1971. A Revision of the Bees of the Genus Andrena of the Western Shields, J.M., Webster, C.R., Storer, A.J., 2008. Short-term community-level response of Hemisphere. Part IV. Scrapteropsis, Xiphandrena and Rhaphandrena. Transactions of arthropods to group selection with seed-tree retention in a northern hardwood the American Entomological Society (1890-) 97:441–520. forest. For. Ecol. Manage. 255 (1), 129–139. LaBerge, W.E., 1973. A Revision of the Bees of the Genus Andrena of the Western Simanonok, M.P., Burkle, L.A., 2019. Nesting success of wood-cavity-nesting bees Hemisphere. Part VI. Subgenus Trachandrena. Transactions of the American declines with increasing time since wildfire. Ecol. Evol. 9 (22), 12436–12445. Entomological Society (1890-) 99:235–371. Smith, C., Weinman, L., Gibbs, J., Winfree, R., Resasco, J., 2019. Specialist foragers in LaBerge, W.E., 1980. A Revision of the Bees of the Genus Andrena of the Western forest bee communities are small, social or emerge early. J. Anim. Ecol. 88 (8), Hemisphere. Part X. Subgenus Andrena. Transactions of the American Entomological 1158–1167. Society (1890-) 106:395–525. Sobek, S., Tscharntke, T., Scherber, C., Schiele, S., Steffan-Dewenter, I., 2009. Canopy vs. LaBerge, W.E., 1985. A Revision of the Bees of the Genus Andrena of the Western understory: Does tree diversity affect bee and wasp communities and their natural Hemisphere. Part XI. Minor Subgenera and Subgeneric Key. Transactions of the enemies across forest strata? For. Ecol. Manage. 258, 609–615. American Entomological Society (1890-) 111:441–567. Steffan-Dewenter, I., Münzenberg, U., Bürger, C., Thies, C., Tscharntke, T., 2002. Scale- Legendre, P., Gallagher, E.D., 2001. Ecologically meaningful transformations for dependent effects of landscape context on three pollinator guilds. Ecology 83 (5), ordination of species data. Oecologia 129 (2), 271–280. 1421–1432. Loehle, C., MacCracken, J.G., Runde, D., Hicks, L., 2002. Forest management at Stockhammer, K.A., 1966. Nesting Habits and Life Cycle of a Sweat Bee, Augochlora pura landscape scales: solving the problems. J. Forest. 100, 25–33. (Hymenoptera: Halictidae). J. Kansas Entomol. Soc. 39, 157–192.

9 K.R. Urban-Mead et al. Forest Ecology and Management 482 (2021) 118903

Stockhammer, K.A., 1967. Some notes on the biology of the blue sweat bee, lasioglossum Westerfelt, P., Weslien, J., Widenfalk, O., 2018. Population patterns in relation to food coeruleum (Apoidea: Halictidae). J. Kansas Entomol. Soc. 40, 177–189. and nesting resource for two cavity-nesting bee species in young boreal forest stands. Trelease, W., 1881. The foliar nectar glands of populus. Bot. Gaz. 6, 284–290. For. Ecol. Manage. 430, 629–638. Ulyshen, M.D., 2011. Arthropod vertical stratification in temperate deciduous forests: Williams, N.M., Crone, E.E., Roulston, T’ai.H., Minckley, R.L., Packer, L., Potts, S.G., Implications for conservation-oriented management. For. Ecol. Manage. 261 (9), 2010. Ecological and life-history traits predict bee species responses to 1479–1489. environmental disturbances. Biol. Conserv. 143 (10), 2280–2291. Ulyshen, M. D., S. Horn, and J. L. Hanula. 2020. Effects of Chinese Privet on Bees and Williams, N.M., Winfree, R., 2013. Local habitat characteristics but not landscape Their Vertical Distribution in Riparian Forests. Forest Science. urbanization drive pollinator visitation and native plant pollination in forest Ulyshen, M.D., Soon, V., Hanula, J.L., 2010. On the vertical distribution of bees in a remnants. Biol. Conserv. 160, 10–18. temperate deciduous forest. Insect Conserv. Divers. 3, 222–228. Winfree, R., Williams, N.M., Dushoff, J., Kremen, C., 2007. Native bees provide insurance USDA, National Agricultural Statistics Service. 2018. 2018 New York cropland data against ongoing honey bee losses. Ecol Lett. 10 (11), 1105–1113. layer. Woodall, C.W., Monleon, V.J., 2007. Sampling Protocol, Estimation, and Analysis Verschuyl, J.P., Hansen, A.J., McWethy, D.B., Sallabanks, R., Hutto, R.L., 2008. Is the Procedures for the Down Woody Materials Indicator of the FIA Program. United effect of forest structure on bird diversity modifiedby forest productivity. Ecol. Appl. States Department of Agriculture Forest Service. 18 (5), 1155–1170.

10