Checklist of bees (Hymenoptera: Apoidea) from small diversified vegetable farms in south-western Montana

Abstract Background Over three years (2013-2015), we sampled bees using nets and bowl traps on four diversified vegetable farms in Gallatin County, Montana, USA, as part of a study evaluating the use of wildflower strips for supporting wild bees and crop pollination services on farmlands (Delphia et al. In prep). We document 202 species and morphospecies from 32 genera within five families, of which 25 species represent the first published state records for Montana. This study increases our overall understanding of the distribution of wild bee species associated with agroecosystems of the northern US Rockies, which is important for efforts aimed at conserving bee biodiversity and supporting sustainable crop pollination systems on farmlands. New information We provide a species list of wild bees associated with diversified farmlands in Montana and increase the number of published bee species records in the state from 374 to at least 399. The list includes new distributional records for 25 wild bee species, including two species that represent considerable expansions of their known ranges, Lasioglossum (Dialictus) clematisellum (Cockerell 1904) with previously published records from New Mexico, Arizona, California and Utah and Melissodes (Eumelissodes) niveus Robertson 1895 which was reported to range from New York to Minnesota and Kansas, south to North Carolina, Alabama and Mississippi.


Introduction
Native bees are important pollinators of wild and cultivated plants in natural habitats and agricultural systems (e.g. Losey and Vaughan 2006, Klein et al. 2007, Garibaldi et al. 2013). In the United States and worldwide, however, bees and other pollinators are experiencing probable declines due to factors such as diseases, pesticides and habitat loss that can reduce floral resources and nesting sites (Biesmeijer et al. 2006, Potts et al. 2010, Cameron et al. 2011, Burkle et al. 2013, Goulson et al. 2015, Koh et al. 2016. However, because we lack a baseline understanding of the bee species that occur in many parts of the US, particularly in certain western regions of the country, concerns regarding the status and trends of wild bees cannot be accurately assessed. Montana's bee fauna is one of the least-studied amongst US states (but see Kuhlman and Burrows 2017, Dolan et al. 2017, Reese et al. 2018; of the few studies that have been conducted in Montana, even fewer have examined wild bees in agricultural systems. Documenting the diversity of bees on farmlands is important for identifying potential crop pollinators, for gauging the potential of farmland habitats to support overall bee diversity on and around farms and for guiding bee conservation measures (Burkle et al. 2017). In addition, the spine of North America, which is the Rocky Mountains, extends from British Columbia to New Mexico and runs through western Montana. This major geographic barrier separates east and west biotas and provides a habitat for alpine/boreal species. Therefore, understanding wild bee distributions in Montana is biogeographically important because its regional species pool likely includes many bee species with typically eastern, western, arctic and southern US ranges that overlap within the state (Gibbs 2010, Koch et al. 2012, Williams et al. 2014, Dolan et al. 2017. A 2017 study, for example, documented 28 bumble bee species in the state, several of which were previously considered to have purely eastern or western North American ranges (Dolan et al. 2017). Another recent survey documented 34 bee species as new state records from just one of Montana's 56 counties (Kuhlman and Burrows 2017). The results from these studies, coupled with Montana's large size (381,000 km ) and diverse ecosystems, many of which are difficult to access, suggest there is much to learn about the wild bee fauna in this region.
We report a checklist of bee species from a three-year study surveying the bee community on diversified vegetable farms in south-western Montana. This is the second study (along with Adhikari et al. In prep) in the state to survey wild bees in agroecosystems, though the habitats (Greater Yellowstone Ecosystem versus Northern Great Plains), type of farming systems (diversified versus highly-simplified) and crops (pollinator-dependent versus windpollinated) differed extensively. This work contributes to our long-term goal of creating a comprehensive bee species list for the state.

Study Sites
This research was conducted at four diversified farms located in south-western Montana USA within a 24 km radius of Bozeman (45.6769°N; 111.0429°W) in Gallatin County (Fig.  1, Table 1). The farms we surveyed are within the eastern end of the broad Gallatin Valley, which is surrounded by five mountain ranges, two of which are nearby: the Bridger Mountains to the northeast and the Gallatin Range to the south. Each farm had approximately 3-7 acres in cultivation each year and grew a variety of crops, including squashes and pumpkins (Cucurbita pepo L.), tomatoes ( Solanum lysopersicum L.), cucumbers (Cucumis sativus L.) and strawberries ( Fragaria x ananassa Duchesne), marketed locally through Community Supported Agriculture (CSA) programmes, farmers' markets, food co-ops and restaurants; two of the farms were certified organic and two followed organic or sustainable practices. Elevations of the farms ranged from 1350-1511 m above sea level. The mean annual precipitation in the area is 469 mm, the mean annual daily high temperature is 12.89°C and the mean annual low temperature is -0.44°C (Western Regional Climate Center 2018; Suppl. material 1). As we were interested in evaluating wildflower strips for supporting bees and crop pollination, our experimental design included planting wildflowers and experimental crop strips from which we sampled bees.  vegetation and time availability. We collected for a total of ca. 23 hours in 2013, 104 hours in 2014 and 85 hours in 2015. In all three years, the same experienced netter (15 years) was paired with a less-experienced netter (≤1 year) for bee collections with both contributing equal amounts of time to collecting; in 2015 the same netter from 2014 assisted with bee collections. Bees were freeze-killed, pinned, and labelled. We also collected bees weekly using yellow, 350-ml Solo bowls filled with soapy water. Six bowls were deployed approximately 6 m apart along each of four, 33-m linear transects (24 bowls per farm) located at different distances from the wildflower strips and left out for approximately six hours during the height of bee activity. Samples were collected into 70% EtOH and later removed from the alcohol, washed, blow-dried, pinned and labelled. We used EstimateS (Colwell 2013) to generate a Chao1 species richness predictor to estimate the "true" number of species present.

Species Identification
We identified bees to the lowest taxonomic level possible using published keys (Table 2) following the classifications of ; specimens were identified to subspecies only when they could be accurately assigned. We used reference specimens from the US National Pollinating Insect Collection to identify a comprehensive subset of the bees collected in this study; these were then taken back to the O'Neill and Burkle Laboratories to use as reference specimens for species identifications and verification of the remaining material.
For genera where no taxonomic literature was available for species-level identifications, we grouped bees that appeared morphologically distinct into morphospecies. We assigned each morphospecies a unique number and the letter "F" for females and "M" for males. However, because we could not reliably associate male and female morphospecies as a Table 2.
List of published keys used for species identification. Genus-level identifications were done using Michener et al. (1994).
Checklist of bees (Hymenoptera: Apoidea) from small diversified vegetable ... single "species" and to avoid inflating species numbers, we included only female morphospecies in this checklist. Species names with aff. ('has affinity with') are also treated as morphospecies.
As females of Agapostemon angelicus Cockerell and Agapostemon texanus Cresson are indistinguishable from one another (Roberts 1972), they have been included with the confirmed males for species counts. Similarly, we could not distinguish amongst females of Hylaeus mesillae (Cockerell) and Hylaeus rudbeckiae (Cockerell and Casad) and they have been included with the males for species counts. For the Chao1 analysis, we have distributed the numbers of females of A. angelicus/texanus and H. mesillae/rudbeckiae according to the proportions of confirmed males of each species for each of the two genera.
Due to a paucity of regional keys and an inability to discern any distinctive characters amongst individuals, except for one morphospecies, bees in the genus Sphecodes were only identified to genus. Similarly, Lasioglossum of the subgenus Evylaeus were identified only to subgenus. Due to time and resource constraints, a randomly-chosen subsample of 66 of the 4,173 Lasioglossum (Dialictus) collected in this study were identified to species.
Voucher specimens will be deposited in the Montana Entomology Collection (MTEC) at Montana State University, Bozeman, MT USA.

Range
To determine whether a species was a new state record, we compared our checklist with other published checklists and literature focused on Montana fauna (O'Neill and Seibert 1996, Fultz 2005, Pearce et al. 2012, Kuhlman and Burrows 2017, Dolan et al. 2017, Reese et al. 2018, as well as a recent unpublished study (Adhikari et al. In prep). These comparisons revealed a subset of species unique to our study; to ensure that these were first records for the state, we reviewed each of these species in the Catalog of Hymenoptera in America north of Mexico (Hurd 1979) for additional Montana records.
Where the catalogue included a record from Montana, we did not conduct further searches for specific localities since our goal was to discover new records for the state. For those not listed as present in the catalogue, we used it to guide further literature searches. We then used published primary literature (see Literature Cited below in checklist notes) to further search for Montana records and determine species ranges. Our search revealed 25 unpublished state records. We also searched DiscoverLife.org; it revealed unpublished records with specific locality information for 5 of the 25 species (Ascher and Pickering 2018; Suppl. material 2), which are indicated below in the checklist notes. For each of the 25 new state records, we also provide information on the closest records reported within the same literature examined for Montana records (see checklist notes).

Protandrena (Pterosarus) piercei (Crawford 1903)
Notes: New species record for Montana (Cockerell 1922; Table 1: Site 1). The closest records reported in Hurd (1979) for this species are from neighbouring Canadian province Alberta and from neighbouring US state North Dakota.

Dufourea marginata (Cresson 1878)
Notes: New species record for Montana (Michener 1951; Table 1: Sites 1, 4). The closest records reported in Hurd (1979) for this species are from neighbouring Canadian province Alberta and from neighbouring state Wyoming.

Discussion
Our study provides information on the wild bee species associated with diverse, smallscale agricultural farmlands in south-western Montana, expands the known distribution of several bee species to Montana and adds to a growing state list.  Reese et al. (2018). Considering that two of these lists are from just one county each (present study and Kuhlman and Burrows 2017) and one (Reese et al. 2018) included three of Montana's 56 counties and that Montana comprises a large, topographically and ecologically diverse region with disparate climatic conditions, this is only a start to the additional surveying that needs to be conducted across the state.
It is not currently possible to accurately compare the number of bee species in Montana to other US states and Canadian provinces, since we are far from a statewide inventory. In addition, few species lists have been published for western US states, though lists have been published for several mid-western and north-eastern US states. The closest, comprehensive, statewide bee list is from Colorado which has 946 species (Scott et al. 2011). Wyoming, Montana's neighbour to the south, has 487 documented bee species with additional species (>150) predicted based on distributional patterns, though faunal assessments are not complete (Lavigne and Tepedino 1976 Our findings support the importance of documenting Montana's bee fauna for understanding the full distributional ranges of the wild bees in North America and for producing more comprehensive regional keys. In addition to wild, native bee species, we also documented four non-native species, including two economically important, commercially-managed species intentionally introduced for crop pollination and two wild, non-managed species accidentally introduced to the US (Klein et al. 2007, Pitts-Singer and Cane 2011, Russo 2016. Managed honey bee, Apis mellifera, colonies were located at one of our sites, as well as nearby sites and were captured in bowl traps and observed visiting our experimental crop strips and wildflower strips. We also captured alfalfa leafcutting bees, Megachile rotundata, in bowl traps and with nets from two non-native plant species, Lotus corniculatus L. and Melilotus officinalis (L.) Lam. Megachile rotundata are managed for alfalfa (Medicago sativa L.) seed production (Pitts-Singer and Cane 2011 in Montana, an important alfalfa seed-producing state (USDA-NASS 2017). Another leafcutting bee Megachile apicalis was recently documented in the literature to occur in Montana (Kuhlman and Burrows 2017) and was netted from E. speciosus in our wildflower strips as well as captured in bowls. Last, the yellow-faced bee Hylaeus leptocephalus was netted from E. speciosus and G. viscosissimum in our flower strips. The impacts (positive and negative) of both intentionally and accidentally introduced non-natives on other bee species (e.g. competition for floral resources and nesting sites and pathogen transmission) and plants (e.g. pollination of native plants, invasive weeds and agricultural crops) require further study (reviewed in Russo 2016).
Our results underestimate the actual bee richness from this study. Though the vast majority of bees were identified to species in our study, the absence of revisionary studies for several genera or subgenera precluded morphospecies sorts for some and, for others, the morphospecies counts may be low because male morphospecies, some of which might not be conspecific with any of the female morphospecies, were not counted. Bees in one genus, Nomada, could only be designated as morphospecies and bee species of Sphecodes and Lasioglossum (subgenus Evylaeus) were classified only to the generic or subgeneric level. In addition, only a fraction of the Lasioglossum in the subgenus Dialictus, which accounted for about 25% of the specimens collected in our study, were identified to species. Bees in the subgenus Dialictus are very abundant in studies using bowl traps (Droege et al. 2010) and the lack of species-level identifications for this group are similar amongst faunistic studies in Montana (e.g. over 22,000 unidentified Dialictus in Kuhlman and Burrows 2017 and over 2,500 unidentified Dialictus in Adhikari et al. In prep). All of these groups require taxonomic work in the western US before species can be fully resolved and will likely contribute dozens of species to a state list for Montana once identified.
Different habitats throughout the state, particularly farmlands versus wildland habitats, are likely to support different suites of bee species. For example, between 33-43% of species were unique to our study when compared to all species (excluding morphospecies) documented in each of two studies conducted in montane wildland habitats in Montana (Kuhlman and Burrows 2017, Reese et al. 2018). However, between 57-66% of bee species in our study were shared with each of those same two studies, indicating considerable overlap with our study and that some bee species can be supported by both types of habitats. When we compared bee species (excluding morphospecies) from all three studies, we found 58 of 195 species (30%) unique to Kuhlman and Burrows (2017), 73 of 226 species (32%) unique to Reese et al. (2018) and 40 of 170 species (24%) unique to our study. All three studies shared 81 species in common, accounting for almost half (48%) of the species in our study, 42% of the species in Kuhlman and Burrows (2017) and 36% in Reese et al. (2018). However, our comparisons did not account for differences in collection methods, sampling effort and geographic area amongst studies, all important determinants of species overlap. More standardised inventories, as well as complete statewide surveys, are needed for more accurate comparisons between studies and habitats.
(Such efforts are hampered by the taxonomic impediment.) In contrast, when we compared our study to another conducted in a highly-simplified, small grains-wheat farming system in north central (Chouteau County) Montana (i.e. the drylands of the Northern Great Plains), we found 73% of bee species were unique to our study compared to those reported by Adhikari et al. (In prep), 27% of bee species were shared (four species are amongst the new state records reported here) and 39 of 85 species (46%) of bee species were unique to Adhikari et al. (In prep). These results suggest the bee communities, which these two agricultural habitats support, are quite different and may be further driven by regional ecosystem diversity. Furthermore, if we compare all four studies, only 28 bee species (16% of the bees in our study) are shared. Again, methodological differences (e.g. sampling intensity, geographic area) between studies make comparisons difficult. Additional surveying is greatly needed in different habitats throughout the state to better understand the basic biology, ecology and distribution of Montana's wild bees, whose importance to natural-and agroecosystems is not fully understood. These types of data are valuable for directing projects aimed at supporting farmland biodiversity and for conserving wild, native bees in general throughout the state.