Plant–Pollinator Networks in Grassland Working Landscapes Reveal Seasonal Shifts in Network Structure and Composition 1 1, 1 1 CAYLA R

Plant–pollinator networks in grassland working landscapes reveal seasonal shifts in network structure and composition 1 1, 1 1 CAYLA R. BENDEL , KATHERINE C. KRAL-O’BRIEN, TORRE J. HOVICK, RYAN F. LIMB, AND 2 JASON P. H ARMON

1Range Science Program, North Dakota State University, 1230 Albrecht Boulevard, 201 Morrill Hall, Fargo, North Dakota 58102USA 2Department of Entomology, North Dakota State University, 1300 Albrecht Boulevard, 202 Hultz Hall, Fargo, North Dakota 58102 USA

Citation: Bendel, C. R., K. C. Kral-O’Brien, T. J. Hovick, R. F. Limb, and J. P. Harmon. 2019. Plant–pollinator networks in grassland working landscapes reveal seasonal shifts in network structure and composition. Ecosphere 10(1):e02569. 10. 1002/ecs2.2569

Abstract. Declines in native bee populations can limit pollination services that support native plant communities and global food production. Mitigating the impacts on pollinators and ecosystems requires conservation actions that promote biodiversity and remain practical for producers. We investigated plant– pollinator interaction networks in working grassland landscapes, managed for cattle production and biodiversity, to advance conservation of pollinators in grazed systems. We compared and plotted interactions at the network level. We then used a regression framework to evaluate the influence of floristic availability on pollinator abundance in our system. Overall, we detected seasonal shifts at the network level, with increased specialization between flowers and bees occurring at the end of the sampling season. Further- more, the response to floristic resources differed between honey bees (Apis mellifera) and native bees. While honey bee abundance increased with exotic floral abundance, native bee abundance showed no relationship with floral abundance and instead was positively associated with floral richness in our system. These findings could be an indication of seasonal shifts in bee activity and interactions with plants or a response to the subset of available resources in grazed systems. These interpretations, along with the detected differ- ence in resource use between honey bees and native bees, suggest foraging preferences differ between these two groups and could influence conservation and management strategies. Furthermore, it demon- strates a need to consider how management practices could influence bee communities differentially across the growing season and suggests conservation actions should promote native floristic resources to benefit native bees.

Key words: grassland ecology; insect conservation; native bee; plant–pollinator interactions; working landscape.

Received 31 July 2018; revised 28 November 2018; accepted 30 November 2018. Corresponding Editor: T’ai Roulston. Copyright: © 2019 The Authors. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. E-mail: [email protected]

INTRODUCTION along with the stability and production of ecosystem services (Hammond 1995). Declines in Human-induced global change has generally biodiversity can reduce the strength and number led to a decline in quality and extent of the of potential species’ interactions and decrease the world’s ecosystems (Solomon et al. 2009, Cardi- ability of an ecosystem to provide services soci- nale et al. 2012). Declines in genetic, species, and ety relies on such as clean water, soil retention, functional biodiversity may be the most promi- and pollination (De Groot et al. 2002, Cardinale nent of these effects (Cardinale et al. 2012, WWF et al. 2012). Thus, biodiversity conservation will 2014). Biodiversity is fundamental for maintain- be critical to maintaining ecosystem services in ing species’ interactions (Cardinale et al. 2012), the future.

❖ www.esajournals.org 1 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL.

Pollination is an economically valuable ecosys- equates to functional redundancy. Therefore, if tem service, but increasing evidence suggests one pollinator goes extinct in the network, pollinators that sustain ecosystem stability and another species will be able to fulfill the func- agricultural production are declining (Kearns tional role in the ecosystem and maintain stabil- et al. 1998, Grixti et al. 2009, Potts et al. 2010), ity. Thus, reduced pollinator diversity reduces although not all pollinators are in decline (e.g., network stability, and increasing plant species Sheffield et al. 2016). Crop pollination has signif- diversity and abundance is essential for pollina- icant economic value as bees are responsible for tor diversity and abundance (Potts et al. 2003, pollinating two-thirds of the world’s crops Hoiss et al. 2015). Additionally, competition and (Williams et al. 2010). While much of the pollina- phenological changes can eliminate plant–polli- tion can be attributed to European honey bees nator interactions before actual individual spe- (Apis mellifera; Morse and Calderone 2000), cies become extinct (Santamarıa et al. 2016) and native pollinators, including bees, are responsible contribute to future species’ extinctions (Tyliana- for an estimated $3.07 billion of food production kis et al. 2010). Therefore, the diversity of polli- in the United States annually (Losey and nators and their interactions with plants should Vaughan 2006). Pollinators also play a key role in be a central tenant of ecosystem conservation maintaining biotic communities in non-agricul- efforts (Carman and Jenkins 2016). tural areas by providing pollination services to Current pollinator research has mainly focused non-crop plants and facilitating gene flow (Potts on managed honey bees (Aizen and Harder et al. 2010, Vanbergen 2013). With pollinator 2009, Pettis and Delaplane 2010, Mogren and declines increasingly recognized as a global Lundgren 2016, Smart et al. 2016). This research conservation concern (Kearns et al. 1998, Potts generally drives policy initiatives, and a honey et al. 2010, Garibaldi et al. 2013), understanding bee centric focus can have indirect negative con- how to maintain plant–pollinator interactions is sequences on other, native bee species (Colla and valuable for crop production, native plant MacIvor 2016). Recent research suggests that the management, and overall the conservation of addition of honey bee hives to a landscape biodiversity (Allen-Wardell et al. 1998). results in native bumble bee declines (Her- One goal of ecosystem management should be bertsson et al. 2016), and honey bees have been promoting native pollinator populations. Pollina- found to negatively impact bumble bee popula- tor abundance and diversity have been positively tions via resource competition (Thomson 2016). correlated with pollination services (Winfree Therefore, specific management strategies benefi- et al. 2015, Orford et al. 2016), but bee diversity cial for the non-native honey bee may not pro- can also bolster ecosystem stability (Potts et al. mote native bee populations. While honey bees 2010, Orford et al. 2016). Native bees have are effective pollinators of certain monoculture evolved with native plants in their respective crops, maintaining diverse native bee communi- biomes (Kearns and Inouye 1997, Kearns et al. ties likely has greater benefit to agricultural 1998), whereas non-native bees, mainly honey yields (Klein et al. 2003, Garibaldi et al. 2015, bees (Apis millifera) in the United States, are often Orford et al. 2016). However, decreases in floral introduced for agricultural purposes and may resource availability and distribution are con- show preference for exotic floral species and alter tributing to native pollinator declines (Potts et al. pollinator community composition (Goulson 2010). Even moderate increases in flowering 2003, Gonzalez-Varo et al. 2013). While honey resources in agricultural landscapes (e.g., alfalfa) bees are numerous and can be effective pollina- could promote pollinator populations while tors for some plants (Huryn 1997), a diverse com- simultaneously promoting biodiversity that sus- munity of bees may be required to maximize tains ecosystem stability and native plant com- productivity within a landscape (Klein et al. munities (Kearns et al. 1998, Ashman et al. 2004, 2003, Orford et al. 2016), and the diversity of Orford et al. 2016). plant–pollinator interactions, or networks, is Our study examined plant–pollinator interac- important for maintaining ecosystem stability tions in grazed working landscapes—ecosystems (Kearns et al. 1998, Ashman et al. 2004). In the- managed for both biological and economic objec- ory, increased pollinator diversity in a network tives (Polasky et al. 2005). Efforts to effectively

❖ www.esajournals.org 2 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL. conserve biodiversity will inevitably need to surrounding area are managed with a combina- incorporate these types of landscapes because tion of fire, mowing, and grazing leases (United the majority of grasslands are in private owner- States Department of Agriculture 2008). ship that rely on livestock production for income We selected eight pastures comprising a vari- (Herrero et al. 2009, Mora and Sale 2011). Our ety of grazing management strategies for use in objectives were to (1) describe plant–pollinator this study. Five of the eight pastures were located interactions in working grassland landscapes at in the SNG and managed by the United States the network level and (2) quantify the relation- Forest Service. The remaining three were located ship between floral availability and bee abun- adjacent to the SNG and managed by North dance. We hypothesized that floral resource Dakota State University. Pastures ranged in size availability would positively influence network- from 54 to 484 ha. All pastures were working level interaction metrics as well as pollinator landscapes managed for cattle production with abundance due to other evidence of bee commu- similar stocking rates, but different cattle man- nities’ responses to floral resources (Ebeling et al. agement strategies were utilized at each site (two 2011, Wray et al. 2014, Goulson et al. 2015). This pastures per strategy). Our initial study was information will be important for understanding designed to test for differences between manage- bee resource use and will improve management ment strategies. However, due to logistical con- that can effectively support native bee communi- straints, treatments were not fully implemented ties in working landscapes. to correspond with the timing of this study. Moreover, management strategies did not influ- METHODS ence floral resources and the butterfly community over the same time period that bee data Site description were collected (Bendel et al. 2018), and we From 2015 to 2016, we evaluated the influence therefore did not evaluate changes in the pollina- of floral availability on pollinator communities tor and floral community across management and foraging in the Sheyenne National Grass- strategies. lands (SNG) and surrounding area located in Richland and Ransom counties, North Dakota, Data collection USA (46.3815° N, 97.2760° W). These areas are Floral visitor surveys.—We sampled the bee located in the tallgrass prairie ecoregion and community from June to August of 2015–2016 characterized by sandy soils and native grasses across three sampling rounds per year (June 10– such as big bluestem (Andropogon gerardii), June 30, July 1–July 20, July 21–August 10). We switchgrass (Panicum virgatum), and prairie cord- used 24, 25 m transects that were randomly grass (Spartina pectinata), but they are heavily distributed within each pasture for a total of 192 invaded by Kentucky bluegrass (Poa pratensis) floral visitor transects. Sampling occurred and smooth brome (Bromus inermis). Common between 09:00 and 17:30 hours on sunny days native forbs include leadplant (Amorpha canes- with temperatures ≥17°C and sustained winds cens), common milkweed (Asclepias syriaca), pas- <15 km/h (Moranz et al. 2012, Popic et al. 2013). ture rose (Rosa arkansana), and meadow anemone We sampled each floral visitor transect for eight (Anemone canadensis). Additionally, invasive forb minutes attempting to collect all flower visitors species such as leafy spurge (Euphorbia esula), within 1 m of either side of the transect. Han- sweet clover (Melilotus spp.), and white clover dling time was not included in the eight minutes. (Trifolium repens) are abundant (USDA 2008). The Every specimen was collected, so we were able region has a temperate climate with cold winters to move forward and backward on the transect and warm, dry summers. The average annual without double sampling. For the focus of this precipitation is 52.6 cm, and average annual tem- study, we targeted bees while collecting, but peratures are 5.5°C (NDAWN 2015). During the used post-sampling identification in the labora- growing season (June–August), the historical tory to refine our analysis to exclusively bees. average temperatures are 19.4°, 22.2°, and In order to identify specific bee–flower interac- 21.1°C, and rainfall totals are 9.8, 8.9, and 5.4 cm, tions, we collected floral visitors using an aerial respectively (NDAWN 2015). The SNG and insect net. We only collected floral visitors

❖ www.esajournals.org 3 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL. touching the reproductive parts of a flower. After software R (version 3.3.1., R Core Team 2016) an individual was netted on a flower, we and the package bipartite (Dormann et al. 2018). recorded the associated flower species each spec- We pooled data at the pasture level for each imen was collected from and euthanized speci- observation period resulting in three plant– mens with ammonium carbonate in a vial. flower interaction webs (Ebeling et al. 2011). We Actively collecting floral visitors from individual generated both quantitative indices as well as flowers provides information about plant–polli- qualitative metrics in the form of interaction nator interactions and more accurately reflects plots to visually evaluate differences in networks use of a resource (Popic et al. 2013). Additionally, across the growing season. we recorded the number and species of flowering We evaluated plant–pollinator interactions stems within 2 m of each 25 m transect (modi- across the growing season by calculating the fi fl ed from Moranz et al. 2014) during each oral- quantitative network-level index H2 for the eight visitor sampling round to make predictions pastures in each sampling period both years. H2 between floral abundance (the total number of assesses the specialization at the network level. flowering stems counted in belt transects), floral In a bee (c = columns) by plant (r = rows) richness (the total number of unique species matrix, let pij be the proportion of the frequency detected in belt transects), and pollinators. This of interactions between a plant (j) and bee species – information was necessary to create plant polli- (i) in respect to the total number of interactions nator networks and foraging preference among (sum of rows [r] and columns [c]) in the network available resources over time. (Bluthgen€ et al. 2006). Bee identification.—We stored specimens in Xr Xc plastic bags indoors at room temperature for H ¼ p ln p the remainder of the field season which helped 2 ij ij i¼1 j¼1 prevent specimen damage. We assigned each fi 0 specimen a unique identi cation code that We present the standardized index H2 which related them to the flower, transect, pasture, ranges from 0 (most generalized) to 1 (most spe- and date they were collected. At the end of the cialized) network (Bluthgen€ et al. 2006). In the field season, we used UNITRON Z10 Stereo index, 0 represents a perfectly nested network microscopes and the key from The Bee Genera of (e.g., progressively specialized bee species visit a North and Central America (Michener et al. 1994) subset of flower species visited by the most gen- to identify specimens down to the lowest taxo- eralized bee), and 1 is the greatest deviation from nomical class practical. We additionally sought this assemblage (e.g., each bee species has a assistance from entomologists and reference col- mutually exclusive interaction with flower spe- lections at USGS Northern Prairie Wildlife cies; Bluthgen€ et al. 2008, Hulsmann€ et al. 2015). 0 Research Center. The index H2 accounts for the frequency of interactions and thus gives a more quantitative analy- Data analysis sis of the pollinator network which can then be Bee and floral richness.—To see broad patterns compared to other networks of varying composi- between floral availability and the bee commu- tion and interactions, instead of just comparing nity, we calculated basic metrics of the floral and the number of links in the network or con- bee communities including abundance, richness, nectance (proportion of realized vs. potential and number of flowering species used by bees links; Bluthgen€ et al. 2006). among sampling rounds (Early–Late). We used To quantify differences in the pollinator net- these data for evaluating differences in plant– works, we used analysis of variance (ANOVA) in 0 pollinator interaction networks from each sam- R (version 3.3.1.; R Core Team 2016). The H2 fol- pling round. lowed a normal distribution, so we did not trans- Network analyses.—Ecological networks encom- form the data. We evaluated changes in the index 0 pass the interactions of species within an ecosys- H2 over sampling periods with year and site as tem and act as a proxy to assess the stability of random factors. For any significant differences an ecosystem (Montoya et al. 2006). We con- (P ≤ 0.05), we used post hoc Tukey tests for pair- ducted all network analyses using the statistical wise comparisons.

❖ www.esajournals.org 4 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL.

Linear regressions.—To quantify relationships the study, but frequently encountered transects between floral resources and bee abundance, we with no individuals. The most abundant species used generalized linear mixed-effect modeling was the honey bee which comprised 647 of the using the lme4 package in R (version 3.3.1.; R total specimens. The second most abundant spe- Core Team 2016). Our response variables were cies at 43 observations was a sweat bee (Lasioglos- honey bee abundance and native bee abundance. sum perpunctatum), and we made 35 observations The a priori floral resource variables (explana- of brown-belted bumble bees (Bombus griseocol- tory variables) were total floral abundance, lis). Overall, bee abundance across sampling native floral abundance, exotic floral abundance, rounds varied between 2015 and 2016, but bee total floral richness, and native floral richness. species richness was the highest in the last round We tested for correlations and removed variables of sampling in both years (Table 1). that were highly correlated (r ≥ 0.60). Exotic flo- We detected 82 different flowering plant spe- ral abundance was correlated with total floral cies on the landscape, 5 of which were exotic. We abundance, but neither variable was highly cor- found that invasion by exotic species such as related with native floral abundance. Total floral leafy spurge, white clover, and sweet clover richness was highly correlated with native floral formed extensive patches in most pastures and richness. Final explanatory variables were exotic accounted for 79% of the total number of flower- floral abundance, native floral abundance, and ing stems over the entire study. During each sam- total floral richness. pling round, exotic species varied in their We conducted regressions using transect-level percentage of total floral abundance at the site data that included site as a random factor. Addi- level (0–100%), with an average of 68% of floral tionally, we used Poisson distributions when cre- abundance compromised of exotic species. Over- ating models for honey bee and native bee all, floral abundance across rounds varied abundance to account for the skewedness in the between 2015 and 2016 (Table 1). However, in count data. During floral visitor surveys, we both years, floral species richness was lowest in occasionally had no bee captures, resulting in the first round and highest in the second round zero count data. We considered any results with of sampling, and the number of flowering species a P-value ≤0.05 to be significant. that bees were collected from generally increased throughout the season (Table 1). RESULTS Network analyses Bee and floral richness From 2015 to 2016, we detected 215 unique We collected 1111 bee specimens and 68 spe- plant–pollinator interactions, and bees made use cies over 153 h of bee sampling. We observed at of only 39 of the 82 flowering plants available least one bee on 171 of 192 transects throughout throughout the sampling period (Table 1, Fig. 1).

Table 1. Total flowering forb abundance, floral richness, floral species bees were collected from, bee abundance, and bee richness by sampling round (1–3), and year collected at the Sheyenne National Grassland and surrounding areas, ND, from 2015 to 2016.

Floral Bee Sampling round Exotic abundance Native abundance Richness No. species used by bees Abundance Richness

1 (June 10–June 30) 30,420 11,819 24 8 100 18 2 (July 1–July 20) 49,694 10,363 42 10 180 12 3 (July 21–August 10) 35,317 14,852 36 21 162 25 2015 Total 115,431 37,034 63† 27† 442 36† 1 (June 10–June 30) 53,291 9406 31 14 294 25 2 (July 1–July 20) 31,756 4635 37 15 204 23 3 (July 21–August 10) 6807 5575 36 16 171 31 2016 Total 91,854 19,616 82† 32† 669 49† † Values are the sum of unique species found that year or unique species used by bees.

❖ www.esajournals.org 5 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL.

Fig. 1. Plant–pollinator networks sampled from 2015 to 2016 in the Sheyenne National Grasslands and surrounding areas. Networks were combined over the two years and plotted for each sampling round. The top bars represent pollinator species and bottom bars represent individual flowers species. A pollinator species is con- nected by a line to a flower if an interaction occurred (i.e., the bee was collected from that flower). For ease of interpretation, only the honey bee (Apis mellifera) is labeled on the graphic, but bee species are listed from left to

❖ www.esajournals.org 6 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL.

(Fig. 1. Continued) right for each round below. Species in bold are non-bee species that collected from the reproductive parts of the flower. Specimens from the families Icheumonidae and Tiphiidae could not be identified to genus. Flower species are labeled as numbers on the bottom of each graphic, with the key for each number found below. Bar width increases as the relative number of interactions between the same bee and flower species increases. Additionally, 0 0 the H2 for each round is included to indicate the level of network interactions. A higher H2 indicates higher network specialization. Plants: 1, Achillea millefolium; 2, Allium cernuum; 3, Amorpha canescens; 4, Anemone canadensis; 5, Ascelpias syriaca; 6, Asclepias stenophylla; 7, Cirsium discolor; 8, Cirsium flodmanii; 9, Dalea candida; 10, Dalea pur- purea; 11, Dalea villosa; 12, Erigeron strigosus; 13, Erysimum capitatum; 14, Euphorbia esula; 15, Glycyrrhiza lepidota; 16, Helenium autumnale; 17, Liatris aspera; 18, Lythrum alatum; 19, Melilotus albus; 20, Melilotus officinalis; 21, Mentha arvensis; 22, Oenothera biennsis; 23, Onosmodium bejariense; 24, Potentilla arguta; 25, Pycnanthemum virginianum; 26, Ratibida columnifera; 27, Rosa arkansana; 28, Rudbeckia hirta; 29, Senecio plattensis; 30, Solidago missouriensis; 31, Spi- raea alba; 32, Symphoricarpos albus; 33, Taraxacum officinale; 34, Teucrium canadense; 35, Tragopogon dubius; 36, Tri- folium pretense; 37, Trifolium repens; 38, Vernonia fasciculata; 39, Zizia aptera. Pollinators: (June 10–June 30): Agapostemon angelicus, Andrena nasonii, Andrena nivalis, Andrena vicina, Andrena wilkella, Apis mellifera, Augochlor- ella aurata, Augochloropsis sumptuosa, Bombus bimaculatus, Bombus borealis, Bombus fervidus, Bombus griseocolis, Bom- bus ternarius, Bombus vagans, Ceratina dupla, Colletes americanus, Halictus confuses, Halictus ligatus, Halictus parallelus, Hoplitis pilosifrons, Hoplitis truncate, Hylaeus affinis, Hylaeus annulatus, Hylaeus mesillae, Ichneumonidae., Lasioglossum spp., Lasioglossum cressonii, Lasioglossum nelumbonis, Lasioglossum paraforbesii, Lasioglossum pectoral, Lasioglossum perpunctatum Lasioglossum pictum, Megachile brevis, Osmia distincta, Vespula vidua. (July 1–July 20): Agapostemon virescens, Andrena wilkella, Andrena ziziae, Apis mellifera, Augochloropsis sumptuosa, Bombus bimaculatus, Bombus borealis, Bombus fervidus, Bombus griseocollis, Bombus griseocollis, Bombus huntii, Bombus rufocinctus, Bombus ternarius, Bombus vagans, Colletes spp., Colletes aberrans, Colletes americanus, Colletes robertsonii, Epeolus spp., Halictus confuses, Halictus ligatus, Hoplitis pilosifrons, Hoplitis truncate, Hylaeus mesillae, Lasioglossum spp., Lasioglossum pectoral, Lasioglossum perpunctatum, Lasioglossum pictum, Melissodes communis, Nomia universitatis. (July 21–August 10): Agapostemon angelicus, Andrena krigiana, Andrena nubecula, Apis mellifera, Augochloropsis sumptuosa, Bombus auricomus, Bombus bimaculatus, Bombus borealis, Bombus fervidus, Bombus griseocollis, Bombus griseocollis, Bombus huntii, Bombus impatiens, Bombus nevadensis, Bombus sandersoni, Bombus ternarius, Bombus vagans, Ceratina dupla, Cerceris spp., Colletes spp., Colletes aberrans, Colletes solidaginis, Ectemnius maculosus, Epeolus spp., Epeolus scutellaris, Halictus confuses, Halictus ligatus, Hylaeus affinis, Lasioglossum spp., Lasioglossum forbesii, Lasioglossum paraforbesii, Lasioglossum pectoral, Lasioglossum perpunctatum, Lasioglossum pictum, Megachile lati- manus,, Melissodes agilis, Melissodes communis, Melissodes desponsa, Melissodes druriella, Melissodes subillata, Melis- sodes trinodis, Philanthus bilunatus, Polistes fuscatus, Tachytes spp., Tiphiidae, Melissodes subillata, Melissodes trinodis, Philanthus bilunatus, Polistes fuscatus, Tachytes,Tiphiidae.

The most frequently used native floral species between sampling rounds combined over both were Flodman’s thistle (Cirsium flodmanii), lead- years. Specialization increased over the field sea- plant, and Missouri goldenrod (Solidago mis- son with the highest specialization occurring at souriensis). Honey bees were highly specialized the end of the summer, representing the most on leafy spurge and white clover in the begin- exclusiveness in interactions between the three ning of the summer (Fig. 1), but then transi- plant–pollinator networks (Fig. 1). However, the tioned to the exotic white and yellow sweet first two rounds of sampling were not signifi- clover during the second sampling round. As cantly different from each other (Fig. 1). plant phenology progressed, we observed an increase in native floral resources which corre- Linear regressions sponded with an increase in native bees and the We found honey bee abundance increased with overall number of flowering species used by bees increasing exotic floral abundance (z = 21.04, in the networks (Fig. 1). df = 182, P < 0.001, r2 = 0.53), but native bee 0 At the network level, specialization (H2 )was abundance did not follow this relationship fi = < = = = 2 = signi cantly different (F2,31 15.94, P 0.001) (z 0.68, df 182, P 0.50, r 0.0002; Fig. 2).

❖ www.esajournals.org 7 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL.

Fig. 2. Mixed-effect linear regressions that compare relationships between bee abundance and resource availability in the Sheyenne National Grasslands and surrounding areas from 2015 to 2016. Each point shows the result from a single transect with some overlapping points, particularly in the bottom graph. The contour plot (shading) helps show the density of points in a given area (regardless of whether they are overlapping or not), with darker shading indicating more points in a given area of the graph. (Top) Honey bee abundance in a transect increased with the abundance of exotic ﬂowers in that transect. (Bottom) Native bee abundance in a transect was positively associated with the richness of all ﬂower species in that transect.

Additionally, honey bee abundance was not pre- DISCUSSION dicted by total floral richness (z = 0.20, df = 182, P = 0.84, r2 = 0.0003). Conversely, we found a Ecosystem degradation coupled with intensive positive relationship between native bee abun- agricultural practices has led to declines in the dance and total floral richness (z = 4.85, df = 182, world’s biodiversity (Briske et al. 2008, Cardinale P ≤ 0.001, r2 = 0.11; Fig. 2). Neither honey bee et al. 2012, WWF 2014) and could threaten the abundance (z = 0.68, df = 182, P = 0.50, r2 = provision of pollination services and food secu- 0.03) nor native bee abundance (z = 1.41, rity (Kearns and Inouye 1997, Kearns et al. 1998, df = 182, P = 0.16, r2 = 0.0001) was predicted by Potts et al. 2010). Reconciling agricultural pro- native floral abundance alone. duction and conservation of biodiversity is

❖ www.esajournals.org 8 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL. crucial to effective ecosystem management. We Our results are most likely a reflection of bee examined plant–pollinator interactions in grass- phenology and niche partitioning between spe- lands actively grazed by livestock to investigate cies (Samnegard et al. 2015, Venjakob et al. plant–pollinator relationships within agroecosys- 2016). Regression analyses investigating floristic tems and found that the network structure and availability and bee abundance complement our composition shifted across the growing season network analyses but reveal that honey bees and with an increase in native bee species later in the native bees are responding to different resources. growing season. Furthermore, we detected differ- Factors such as size and sociality can drive tem- ences in native bee and honey bee responses to poral variation in bee species’ presence (Ogilvie floral resources. Native bee abundance increased and Forrest 2017). We found that the social honey as total floral richness increased, whereas honey bee increased with exotic floral abundance, an bee abundance increased as exotic floral abun- indication of total floral abundance. Apis species dance increased. Our results are similar to other commonly focus on abundant and dense studies that have reported bottom-up effects of resources, whereas solitary bee species favor flo- resource diversity on pollinators in grazed sys- ral richness (Ebeling et al. 2011, Thomson 2016). tems (Orford et al. 2016) and differences in honey Thus, it is unsurprising that honey bee abun- bee and native bee foraging preferences (Otto dance was predicted by areas dominated by leafy et al. 2017). Together, these results suggest that spurge and sweet clover, two exotic species that bee species are using resources differentially in form dense patches on the landscape compared working landscapes, resulting in the spatial-tem- to native species (Lym and Kirby 1987, Wolf poral fluctuations observed in the plant–pollina- et al. 2003). Conversely, solitary bees may tor networks in this study and elsewhere (see respond to resources at a smaller scale (Gath- Lopez et al. 2017). mann and Tscharntke 2002). For example, spe- Honey bees were a prominent part of plant– cialist bees in the northeastern United States may pollinator networks we observed. North Dakota rely on just one host–plant or nectar source and has the most commercial honey bee colonies in thus have a limited period to be observed (Fow- the country, with several hundred colonies resid- ler 2016). This could explain our finding between ing in our study area alone (USDA 2016). native bee abundance and total floral richness as Nonetheless, plant–pollinator networks for both well as the increase in native bee species seen at honey bees and native bees changed over the the end of the growing season when floral rich- season, and floral resource availability, both ness was at its highest. However, it is important abundance and richness, appears to be influenc- to point out that we did not further categorize ing pollinator abundance. Phenology of floral native bees as solitary vs. social or generalist vs. resources, along with native and exotic bee phe- specialist. All of these variations in resource use nology and biology, likely influenced our net- and network structure can fluctuate between sea- work results. Life history traits and evolutionary sons (Geslin et al. 2016) and be altered by land relationships with plants can influence the timing use and climatic changes (Ogilvie and Forrest of plant–pollinator interactions, all of which can 2017). Therefore, it will be important to conduct be altered by land use (Geslin et al. 2016, Ogilvie longer term studies of plant–pollinator networks and Forrest 2017). Furthermore, grazed systems to best inform conservation in modern-day work- in this study may only offer a subset of floral ing landscapes. resources due to historical land management The observed plant–pollinator network shifts affecting floral resource availability and pollina- in our study may be the result of the limited tor abundance. Generally, increases in floral resource availability in moderately grazed pas- diversity can increase pollinator diversity (Potts tures or phenological changes in resource avail- et al. 2010, Ebeling et al. 2011, Venjakob et al. ability. Grazing can impact bee communities 2016), but the identity of those floral resources indirectly through changes to the floral commu- may be of importance for managing particular nity (Lazaro et al. 2016). Often, grazing reduces groups of bees (Sutter et al. 2017), especially in the availability of floral resources, unlike fire working landscapes. which can increase floral density (Moranz et al.

❖ www.esajournals.org 9 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL.

2014). Grazing has been shown to be compatible (Tscharntke et al. 2005). Overall, management with pollinator conservation (Potts et al. 2006, practices which emphasize the spatial-temporal Williams et al. 2010, Moranz et al. 2012), but distribution of key floristic resources will not landscapes impacted by humans can be only bolster pollinator diversity but the services resource-limiting during certain times of the year they provide (Frund€ et al. 2013). and require bees to travel long distances (Olsson We recognize that our study has limitations as et al. 2015). In particular, early season grazing a consequence of being conducted over two vari- has been shown to be harmful to bee communi- able growing seasons and other challenges of ties because fewer resources are blooming during conducting research in working landscapes. A this time (Lazaro et al. 2016). We observed simi- study greater than two years may help to delin- lar patterns among pastures used in this eate interactions further. Additionally, we did research. In general, fewer flowering species not differentiate between floral resources being were available early in the season, but floral den- used for nectar or pollen. Often species have sity was relatively high due to exotic species. more nectar than pollen hosts, and some flower This is one explanation for the simplified net- species may only provide one or the other work seen in our first sampling period. (Robertson 1925, Baude et al. 2016), influencing Plant–pollinator networks may also change species’ resource use. Finally, due to much of the throughout the growing season due to resource landscape being leased to private ranchers, we availability in adjacent landscape types. Both had insufficient knowledge of current and his- native bees and honey bees may utilize crops toric land use practices to make connections grown in the surrounding area (Garibaldi et al. between management practices and floristic 2013) such as edible beans and peas, sunflower, resource availability. Nevertheless, this research and soybeans (North Dakota Agricultural Statis- contributes to the limited understanding of tics 2017). Additionally, honey bees may use native plant–pollinator networks in grassland wooded areas to collect resin for propolis later in working landscapes and presents a very realistic the growing season (Simone-Finstrom and Spi- outlook of how pollinator communities interact vak 2010). However, bees will still need access to with the floral resources that remain after dec- semi-natural habitat for supplemental resources ades of intense land use. within a reasonable foraging distance (Olsson While the conservation of biodiversity is et al. 2015, Senapathi et al. 2016). Additionally, imperative to supporting ecosystem services, pollinator communities need these resources to responses which forego production entirely are be temporally distributed (Leong et al. 2016). We impractical. Rather, management practices which found that floral availability was predictive of can simultaneously achieve both, need to be the bee community but differed between native researched and applied. In the case of pollina- bees and honey bees. Entire bee communities tors, the grassland ecosystems that provide the may shift throughout the season as a result of bulk of their resources are also essential land- niche partitioning among the resources available scapes for agriculture and livestock production. (Venjakob et al. 2016, Lopez et al. 2017), but a We investigated plant–pollinator networks in suite of flower species should be available to sup- grassland working landscapes grazed by cattle port a diverse pollinator community throughout and found that floral resource availability (abun- the season and across the landscape (Samnegard dance and richness) was influencing plant–polli- et al. 2015, Venjakob et al. 2016). Other research- nator networks and bee abundance throughout ers found that rarer bees used a subset of the summer. By managing for diverse spatial- resources from more abundant native bees and temporal floristic resources, the potential to sup- honey bees (Sutter et al. 2017). Managing for port native bee species across the entire growing these key resources can result in increases in total season can increase. Developing innovative man- bee abundance (Sutter et al. 2017), but manage- agement practices which can provide this floris- ment strategies would undoubtedly change for tic diversity will be imperative to conserving social vs. solitary bee species due to differences native bee species and industries which support in dispersal range and communication abilities global food production.

❖ www.esajournals.org 10 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL.

ACKNOWLEDGMENTS reveals unsuccessful foraging of native bees in dis- turbed habitats. Biological Conservation 202:110– We would like to thank the U.S. Forest Service for 118. permission to sample pastures within the Sheyenne Colla, S. R., and J. S. MacIvor. 2016. Questioning public National Grasslands. We also thank the North Dakota perception, conservation policy, and recovery State University Range Science department for assist- actions for honeybees in North America. Conserva- ing with the prescribed burns necessary for the experi- tion Biology 31:1201–1204. mental design, and we acknowledge Sam O’Dell and De Groot, R. S., M. A. Wilson, and R. M. Boumans. Clint Otto with Northern Prairie Wildlife Research 2002. A typology for the classification, description Center for their aid and resources in identifying bee and valuation of ecosystem functions, goods and specimens. Thank you to three anonymous reviewers services. Ecological Economics 41:393–408. for providing helpful feedback. Finally, we thank the Dormann, C.F., J. Fruend and B. Gruber. 2018. Package North Dakota State University Agricultural Experi- ‘bipartite’. Version 2.11. http://r.adu.org.za/web/ ment Station for funding this project. packages/bipartite/bipartite.pdf Ebeling, A., A. M. Klein, and T. Tscharntke. 2011. LITERATURE CITED Plant–flower visitor interaction webs: Temporal stability and pollinator specialization increases Aizen, M. A., and L. D. Harder. 2009. The global stock along an experimental plant diversity gradient. – of domesticated honey bees is growing slower than Basic and Applied Ecology 12:300 309. agricultural demand for pollination. Current Biol- Fowler, J. 2016. Specialist bees of the Northeast: host ogy 19:915–918. plants and habitat conservation. Northeastern Nat- – Allen-Wardell, G., et al. 1998. The potential conse- uralist 23:305 320. € quences of pollinator declines on the conservation Frund, J., C. F. Dormann, A. Holzschuh, and T. of biodiversity and stability of crop yields. Conser- Tscharntke. 2013. Bee diversity effects on pollina- vation Biology 12:8–17. tion depend on functional complementarity and – Ashman, T. L., et al. 2004. Pollen limitation of plant niche shifts. Ecology 94:2042 2054. reproduction: ecological and evolutionary causes Garibaldi, L. A., I. Bartomeus, R. Bommarco, A. M. and consequences. Ecology 85:2408–2421. Klein, S. A. Cunningham, M. A. Aizen, and M. fl Baude, M., W. E. Kunin, N. D. Boatman, S. Conyers, N. Woyciechowski. 2015. Trait matching of ower vis- Davies, M. A. Gillespie, R. D. Morton, S. M. Smart, itors and crops predicts fruit set better than trait – and J. Memmott. 2016. Historical nectar assessment diversity. Journal of Applied Ecology 52:1436 reveals the fall and rise of floral resources in Bri- 1444. tain. Nature 530:85–88. Garibaldi, L. A., et al. 2013. Wild pollinators enhance Bendel, C. R., T. J. Hovick, R. F. Limb, and J. P. Har- fruit set of crops regardless of honey bee abun- – mon. 2018. Variation in grazing management prac- dance. Science 339:1608 1611. tices supports diverse butterfly communities across Gathmann, A., and T. Tscharntke. 2002. Foraging grassland working landscapes. Journal of Insect ranges of solitary bees. Journal of Animal Ecology – Conservation 22:99–111. 71:757 764. Bluthgen,€ N., J. Frund,€ D. P. Vazquez, and F. Menzel. Geslin, B., M. Oddie, M. Folschweiller, G. Legras, C. 2008. What do interaction network metrics tell us L. Seymour, F. F. Van Veen, and E. Thebault. 2016. fl about specialization and biological traits? Ecology Spatiotemporal changes in ying insect abundance 89:3387–3399. and their functional diversity as a function of dis- fl Bluthgen,€ N., F. Menzel, and N. Bluthgen.€ 2006. Mea- tance to natural habitats in a mass owering crop. – suring specialization in species interaction net- Agriculture, Ecosystems & Environment 229:21 29. works. BMC Ecology 6:1–12. Gonzalez-Varo, J. P., J. C. Biesmeijer, R. Bommarco, S. Briske, D. D., J. D. Derner, J. R. Brown, S. D. Fuhlen- G. Potts, O. Schweiger, H. G. Smith, I. Steffan- € dorf, W. R. Teague, K. M. Havstad, R. L. Gillen, A. Dewenter, H. Szentgyorgyi, M. Woyciechowski, J. Ash, and W. D. Willms. 2008. Rotational grazing and M. Vila. 2013. Combined effects of global on rangelands: reconciliation of perception and change pressures on animal-mediated pollination. – experimental evidence. Rangeland Ecology & Man- Trends in Ecology & Evolution 28:524 530. agement 61:3–17. Goulson, D. 2003. Effects of introduced bees on native Cardinale, B. J., et al. 2012. Biodiversity loss and its ecosystems. Annual Review of Ecology, Evolution, – impact on humanity. Nature 486:59–67. and Systematics 34:1 26. ı Carman, K., and D. G. Jenkins. 2016. Comparing Goulson, D., E. Nicholls, C. Bot as, and E. L. Rotheray. diversity to flower-bee interaction networks 2015. Bee declines driven by combined stress from

❖ www.esajournals.org 11 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL.

parasites, pesticides, and lack of flowers. Science Losey, J. E., and M. Vaughan. 2006. The economic 347:1255957. value of ecological services provided by insects. Grixti, J. C., L. T. Wong, S. A. Cameron, and C. Favret. BioScience 56:311–323. 2009. Decline of bumble bees (Bombus) in the Lym, R. G., and D. R. Kirby. 1987. Cattle foraging North American Midwest. Biological Conservation behavior in leafy spurge (Euphorbia esula)-infested 142:75–84. rangeland. Weed Technology 1:314–318. Hammond, P. C. 1995. Conservation of biodiversity in Michener, C. D., R. J. McGinley, and B. N. Danforth. native prairie communities in the United States. 1994. The Bee Genera of North and Central Amer- Journal of the Kansas Entomological Society 68:1– ica (Hymenoptera: Apoidea). Smithsonian Institu- 6. tion Press, Washington, DC, USA. Herbertsson, L., S. A. Lindstrom,€ M. Rundlof,€ R. Bom- Mogren, C. L., and J. G. Lundgren. 2016. Neonicoti- marco, and H. G. Smith. 2016. Competition noid-contaminated pollinator strips adjacent to between managed honeybees and wild bumble- cropland reduce honey bee nutritional status. Sci- bees depends on landscape context. Basic and entific Reports 6:29608. Applied Ecology 17:609–616. Montoya, J. M., S. L. Pimm, and R. V. Sole. 2006. Eco- Herrero, M., P. K. Thornton, P. Gerber, and R. S. Reid. logical networks and their fragility. Nature 2009. Livestock, livelihoods and the environment: 442:259–264. understanding the trade-offs. Current Opinion in Mora, C., and P. F. Sale. 2011. Ongoing global biodiver- Environmental Sustainability 1:111–120. sity loss and the need to move beyond protected Hoiss, B., J. Krauss, and I. Steffan-Dewenter. 2015. areas: a review of the technical and practical short- Interactive effects of elevation, species richness and comings of protected areas on land and sea. Marine extreme climatic events on plant-pollinator net- Ecology Progress Series 434:251–266. works. Global Change Biology 21:4086–4097. Moranz, R. A., D. M. Debinski, D. A. McGranahan, D. Hulsmann,€ M., H. von Wehrden, A. M. Klein, and S. M. Engle, and J. R. Miller. 2012. Untangling the D. Leonhardt. 2015. Plant diversity and composi- effects of fire, grazing, and land-use legacies on tion compensate for negative effects of urbanization grassland butterfly communities. Biodiversity and on foraging bumble bees. Apidologie 46:760–770. Conservation 21:2719–2746. Huryn, V. M. B. 1997. Ecological impacts of introduced Moranz, R. A., S. D. Fuhlendorf, and D. M. Engle. honey bees. Quarterly Review of Biology 72:275– 2014. Making sense of the prairie butterfly para- 297. dox: the effect of grazing, time since fire, and sam- Kearns, C. A., and D. W. Inouye. 1997. Pollinators, pling period on regal fritillary abundance. flowering plants, and conservation biology. BioS- Biological Conservation 173:32–41. cience 47:297–307. Morse, R. A., and N. W. Calderone. 2000. The value of Kearns, C. A., D. W. Inouye, and N. M. Waser. 1998. honey bees as pollinators of U.S. crops in 2000. Bee Endangered mutualisms: the conservation of plant- Culture 128:1–15. pollinator interactions. Annual Review of Ecology North Dakota Agricultural Statistics. 2017. North and Systematics 29:83–112. Dakota Agriculture. United States Department of Klein, A. M., I. Steffan–Dewenter, and T. Tscharntke. Agriculture, Agricultural Statistics Number 86. 2003. Fruit set of highland coffee increases with the Retrieved from https://www.nass.usda.gov/Statis diversity of pollinating bees. Proceedings of the tics_by_State/North_Dakota/Publications/Annual_ Royal Society of London B: Biological Sciences Statistical_Bulletin/2017/ND%202017%20Annual% 270:955–961. 20Bulletin.pdf Lazaro, A., T. Tscheulin, J. Devalez, G. Nakas, and T. North Dakota Agricultural Weather Network (NDAWN). Petanidou. 2016. Effects of grazing intensity on pol- 2015. Retrieved from https://ndawn.ndsu.nodak. linator abundance and diversity, and on pollination edu/current.html services. Ecological Entomology 41:400–412. Ogilvie, J. E., and J. R. Forrest. 2017. Interactions Leong, M., L. C. Ponisio, C. Kremen, R. W. Thorp, and between bee foraging and floral resource phenol- G. K. Roderick. 2016. Temporal dynamics influ- ogy shape bee populations and communities. Cur- enced by global change: bee community phenology rent Opinion in Insect Science 21:75–82. in urban, agricultural, and natural landscapes. Glo- Olsson, O., A. Bolin, H. G. Smith, and E. V. Lonsdorf. bal Change Biology 22:1046–1053. 2015. Modeling pollinating bee visitation rates in Lopez, D. N., P. A. Camus, N. Valdivia, and S. A. heterogeneous landscapes from foraging theory. Estay. 2017. High temporal variability in the occur- Ecological Modelling 316:133–143. rence of consumer-resource interactions in ecologi- Orford, K. A., P. J. Murray, I. P. Vaughan, and J. Mem- cal networks. Oikos 126:1699–1707. mott. 2016. Modest enhancements to conventional

❖ www.esajournals.org 12 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL.

grassland diversity improve the provision of Greene and B. terricola Kirby (Hymenoptera: Api- pollination services. Journal of Applied Ecology dae) with implications for conservation status 53:906–915. assessments. Journal of Insect Conservation 20: Otto, C. R., S. O’Dell, R. B. Bryant, N. H. Euliss, R. M. 189–199. Bush, and M. D. Smart. 2017. Using publicly avail- Simone-Finstrom, M., and M. Spivak. 2010. Propolis able data to quantify plant–pollinator interactions and bee health: the natural history and significant and evaluate conservation seeding mixes in the of resin use by honey bees. Apidologie 41:295–311. northern Great Plains. Environmental Entomology Smart, M. D., J. S. Pettis, N. Euliss, and M. S. Spivak. 46:565–578. 2016. Land use in the Northern Great Plains region Pettis, J. S., and K. S. Delaplane. 2010. Coordinated of the US influences the survival and productivity responses to honey bee decline in the USA. Api- of honey bee colonies. Agriculture, Ecosystems & dologie 41:256–263. Environment 230:139–149. Polasky, S., E. Nelson, E. Lonsdorf, P. Fackler, and A. Solomon, S., G. K. Plattner, R. Knutti, and P. Friedling- Starfield. 2005. Conserving species in a working stein. 2009. Irreversible climate change due to car- landscape: land use with biological and economic bon dioxide emissions. Proceedings of the National objectives. Ecological Applications 15:1387–1401. Academy of Sciences 106:1704–1709. Popic, T. J., G. M. Wardle, and Y. C. Davila. 2013. Sutter, L., P. Jeanneret, A. M. Bartual, G. Bocci, and M. Flower-visitor networks only partially predict the Albrecht. 2017. Enhancing plant diversity in agri- function of pollen transport by bees. Austral Ecol- cultural landscapes promotes both rare bees and ogy 38:76–86. dominant crop-pollinating bees through comple- Potts, S. G., J. C. Biesmeijer, C. Kremen, P. Neumann, mentary increase in key floral resources. Journal of O. Schweiger, and W. E. Kunin. 2010. Global polli- Applied Ecology 54:1856–1864. nator declines: trends, impacts and drivers. Trends Thomson, D. M. 2016. Local bumble bee decline linked in Ecology & Evolution 25:345–353. to recovery of honey bees, drought effects on floral Potts, S. G., T. Petanidou, S. Roberts, C. O’Toole, A. resources. Ecology Letters 19:1247–1255. Hulbert, and P. Willmer. 2006. Plant-pollinator bio- Tscharntke, T., A. M. Klein, A. Kruess, I. Steffan- diversity and pollination services in a complex Dewenter, and C. Thies. 2005. Landscape Mediterranean landscape. Biological Conservation perspectives on agricultural intensification and bio- 129:519–529. diversity–ecosystem service management. Ecology Potts, S. G., B. Vulliamy, A. Dafni, G. Ne’eman, and P. Letters 8:857–874. Willmer. 2003. Linking bees and flowers: How do Tylianakis, J. M., E. Laliberte, A. Nielsen, and J. Bas- floral communities structure pollinator communi- compte. 2010. Conservation of species interaction ties? Ecology 84:2628–2642. networks. Biological Conservation 143:2270–2279. R Core Team. 2016. R: a language and environment for United States Department of Agriculture (UDSA). statistical computing. R Foundation for Statistical 2008. Sheyenne National Grasslands. Retrieved Computing, Vienna, Austria. from http://www.fs.usda.gov/recarea/dpg/recarea/? Robertson, C. 1925. Heterotropic bees. Ecology 6:412–436. recid=79470 Samnegard, U., P. A. Hamback,€ C. Eardley, S. Nemo- United States Department of Agriculture (UDSA). missa, and K. Hylander. 2015. Turnover in bee spe- 2016. Honey. Retrieved from http://usda.mannlib.c cies composition and functional trait distributions ornell.edu/usda/current/Hone/Hone-03-22-2016.pdf between seasons in a tropical agricultural land- Vanbergen, A. J., and and the Insect Pollination Initia- scape. Agriculture, Ecosystems & Environment tive. 2013. Threats to an ecosystem service: pres- 211:185–194. sures on pollinators. Frontiers in Ecology and the Santamarıa, S., J. Galeano, J. M. Pator, and M. Mendez. Environment 11:251–259. 2016. Removing interactions, rather than species, Venjakob, C., A. M. Klein, A. Ebeling, T. Tscharntke, casts doubt on robustness of pollination networks. and C. Scherber. 2016. Plant diversity increases Oikos 125:526–534. spatio-temporal niche complementarity in plant- Senapathi, D., M. A. Goddard, W. E. Kunin, and K. C. pollinator interactions. Ecology and Evolution 6: Baldock. 2016. Landscape impacts on pollinator 2249–2261. communities in temperate systems: evidence and Williams, N. M., E. E. Crone, T. H. Roulston, R. L. knowledge gaps. Functional Ecology 31:26–37. Minckley, L. Packer, and S. G. Potts. 2010. Ecologi- Sheffield, C. S., L. Richardson, S. Cannings, H. Ngo, J. cal and life-history traits predict bee species Heron, and P. H. Williams. 2016. Biogeography responses to environmental disturbances. Biologi- and designatable units of Bombus occidentalis cal Conservation 143:2280–2291.

❖ www.esajournals.org 13 January 2019 ❖ Volume 10(1) ❖ Article e02569 BENDEL ET AL.

Winfree, R., J. W. Fox, N. M. Williams, J. R. Reilly, and D. Wray, J. C., L. A. Neame, and E. Elle. 2014. Floral P. Cariveau. 2015. Abundance of common species, resources, body size, and surrounding landscape not species richness, drives delivery of a real-world inﬂuence bee community assemblages in oak- ecosystem service. Ecology Letters 18:626–635. savannah fragments. Ecological Entomology 39: Wolf, J. J., S. W. Beatty, and G. Carey. 2003. Invasion by 83–93. sweet clover (Melilotus) in montane grasslands, WWF. 2014. Living Planet Report 2014: summary. Rocky Mountain National Park. Annals of the [McLellan, R., Iyengar, L., Jeffries, B. and N. Oerle- Association of American Geographers 93:531–543. mans (Eds)]. WWF, Gland, Switzerland.

❖ www.esajournals.org 14 January 2019 ❖ Volume 10(1) ❖ Article e02569