Oecologia (2009) 161:813–823 DOI 10.1007/s00442-009-1429-3

COMMUNITY ECOLOGY - ORIGINAL PAPER

Reproduction of () and diversity of its bee community in a fragmented landscape

Malinda W. Slagle · Stephen D. Hendrix

Received: 5 August 2008 / Accepted: 27 July 2009 / Published online: 26 August 2009 © Springer-Verlag 2009

Abstract Loss of pollinators due to habitat frag- eVect on the diversity and abundance of bees visiting A. mentation often results in negative eVects on repro- canescens, indicating community-level characteristics can duction, but few studies have simultaneously examined inXuence the bee community visiting any one species. Site variation in the bee community, site characteristics and size, a common predictor of plant reproduction in frag- plant community characteristics to evaluate their relative mented habitats did not contribute to any models of fruit set eVects on plant reproduction in a fragmented habitat. We and was only marginally related to bee diversity one year. examined the reproduction of a common tallgrass quintilis, one of the three oligolectic bee species forb, Amorpha canescens (Fabaceae), in large (>40 ha) and associated with A. canescens, was abundant at all sites, small (<2 ha) prairie remnants in Iowa and Minnesota in suggesting it has not been signiWcantly aVected by frag- relation to the diversity and abundance of its bee visitors, mentation. Our results show that the diversity of bees plant population size, and species density of the forb Xow- visiting A. canescens is important for maintaining fruit set ering community. We found signiWcant positive eVects of and that bee visitation is still suYcient for at least some the diversity of bees visiting A. canescens on percent fruit fruit set in all populations, suggesting these small remnants set at a site in both years of the study and in 2002 an addi- act as Xoral resource oases for bees in landscapes often tional signiWcant positive eVect of plant species density. dominated by agriculture. Abundance of bees visiting A. canescens had a signiWcant positive eVect on percent fruit set in 2002, but was only Keywords Fruit set · Bee visitors · marginally signiWcant in 2003. In 2003 but not 2002, the Habitat fragmentation · Prairie · Oligolectic bees plant species density at the sites had a signiWcant negative

Introduction Communicated by JeV Conner. The majority of the world’s 220,000 angiosperm species in Electronic supplementary material The online version of this habitats as diverse as Neotropical lowland forests and tall- article (doi:10.1007/s00442-009-1429-3) contains supplementary material, which is available to authorized users. grass depend on insect pollinators for sexual repro- duction (Buchmann and Nabhan 1996; Hendrix and Kyhl M. W. Slagle · S. D. Hendrix (&) 2000). However, wild pollinator diversity, richness, and Department of Biology, University of Iowa, abundance are thought to be threatened by habitat fragmen- Iowa City, IA 52242, USA W e-mail: [email protected] tation and destruction resulting from agricultural intensi - cation and other land use changes (Banaszak 1992; Kremen Present Address: et al. 2007). Loss of wild pollinators may have implications M. W. Slagle for agriculture because they are potential pollinators of crop Litzsinger Road Ecology Center, Missouri Botanical Garden, 9711 Litzsinger Road, typically pollinated by honey bees that have been Saint Louis, MO 63124, USA negatively aVected by colony collapse disorder, Varroa e-mail: [email protected] mites, and other diseases (Hayes 2007; Norton 2007). 123 814 Oecologia (2009) 161:813–823

Despite the general consensus that fragmentation is populations (Molano-Flores and Hendrix 1999) and may be harmful to many native plant and pollinator species, the less able to sustain pollinators, leading to reduced repro- responses of bee communities to fragmentation and the duction for some prairie forbs compared to populations in eVects on the plants they visit and pollinate remain under large remnants (>40 ha) (Lewis 1999; Molano-Flores and active discussion and investigation (Cane 2001; Cane et al. Hendrix 1999; Hendrix and Kyhl 2000). However, the rela- 2006). Of particular interest are the relationships between tive eVects of reduced size and other changes in plant and plant reproduction and reduced pollinator diversity (Klein pollinator communities potentially resulting from fragmen- et al. 2003, 2008), decreased site size (Jennersten 1988; tation of Iowa prairies are unknown. Aizen and Feinsinger 1994; Ågren 1996; Morgan 1999; In this study we measured fruit set of a common prairie Brys et al. 2004), and altered plant population (Richards forb, Amorpha canescens Pursh. (lead plant, Fabaceae), in et al. 1999; Matsumura and Washitani 2000) or habitat scattered remnants in northwest Iowa and southwest Min- patch characteristics (Tscharntke et al. 1998; SteVan- nesota while simultaneously sampling its bee community Dewenter and Tscharntke 1999). EVects of fragmentation and measuring Xoral resource availability of A. canescens on pollinator diversity and abundance should in turn be and other plants at the remnants. We tested and compared examined by studying relationships to site size and Xoral as possible correlates of fruit set critical factors associated diversity (Heithaus 1974; Potts et al. 2003). However, with the fragmentation process including remnant size unraveling the complexity of plant-pollinator interactions (large or small), population size of A. canescens, species in fragmented landscapes is diYcult because few studies diversity and abundance of bees visiting A. canescens, and simultaneously quantify plant reproduction in fragmented plant species density, a measure of species richness at a landscapes and the multiple factors that can potentially remnant. We also examined for eVects of remnant size and aVect it. Such studies are needed to allow direct comparison measures of Xoral resources on diversity and abundance of of predictors of plant reproduction such as population size bees visiting A. canescens. In addition, we examined year- or remnant size with other potential predictors such as to-year variation in the inXuence of these factors on repro- pollinator diversity or abundance. duction in A. canescens because temporal variation in bee The reproduction of bee-visited plants in fragmented communities (Roubik 2001) and plant reproduction landscapes is also complicated by the degree of specializa- (Molano-Flores and Hendrix 1999) can be great, particu- tion of the bees. Plants dependent on oligolectic bees that larly in small habitat fragments. forage on a narrow taxonomic classiWcation such as a or subfamily of plants (Cane and Sipes 2006) may be par- ticularly vulnerable to the eVects of habitat destruction and Materials and methods fragmentation (Buchmann and Nabhan 1996), which may reduce the abundance (Goulson and Darvill 2004) and Study plant genetic diversity (Packer et al. 2005) of oligolectic bees because they are restricted to foraging sites containing their Amorpha canescens is a perennial semi-shrub commonly host plants (Buchmann and Nabhan 1996). In contrast, the found in remnant prairies and savannas throughout the tall- reproduction of plants dependent on polylectic bees, which grass prairie region from Indiana west to Texas and from visit many plant species in diVerent families, may be rela- Texas north to Manitoba (McGregor et al. 1986). Ramets of tively buVered against some of the negative eVects of habi- A. canescens that are presumably from the same rootstock tat fragmentation on pollinator richness (Waser et al. 1996) often occur as closely associated groups or clumps scat- because functional redundancy should allow for high levels tered up to 10 m apart in Iowa’s prairies (McGregor et al. of pollination service with relatively low levels of species 1986; S. Hendrix personal observation). The root system richness (Gamfeldt et al. 2008). Thus, understanding the can penetrate the soil up to a depth of 5 m with lateral eVects of fragmentation on a plant species’ reproduction spread beginning 1 m underground and extending outward is likely to require detailed knowledge of its pollinator radially for 1 m or more (Weaver and Fitzpatrick 1934). community. The inXorescences of A. canescens are compound racemes In Iowa the destruction of 99.9% (11.9 million ha) of the that range from one to 30 branches and contain up to 3,000 pre-settlement tallgrass prairie by about 1900 (Steinauer Xowers. Flowering begins in late June in northwest Iowa and Collins 1996) has created a system of prairie fragments and southwest Minnesota and continues through early which can be used to study the relationships between native August (Lewis 1999). Each Xower can give rise to a single- plant reproduction, pollinator communities, and plant popu- seeded indehiscent pod; fruits mature by early September in lation characteristics. These small remnants (<2 ha) typi- northwest Iowa and southwest Minnesota. cally contain relatively low species richness of Xoral The primary pollinators of A. canescens are solitary bees resources (Leach and Givnish 1996) and/or small plant that make up the majority of its - and nectar-feeding 123 Oecologia (2009) 161:813–823 815 visitors (Robertson 1928; Parrish and Bazzaz 1979; M. Slagle, Table 1 Location, area, and size of 14 sites in northwestern Iowa and personal observation). Adult syrphid Xies are also frequent southwestern Minnesota, USA visitors, but because of the small amount of pollen they Site name County State Area Site size carry relative to bee visitors, they are generally considered (ha) as non-pollinating visitors (Robertson 1928; M. Slagle per- Cayler Prairie [A] Dickinson IA 65.0 Large (3) sonal observation). Therefore, we focused on bee visitors. V A. canescens is an obligate outcrosser with Xowers that are F. Ha ner Prairie [B] Dickinson IA 45.0 Large (3) protogynous (Cruden 1977). Anderson Prairie [C] Emmett IA 80.0 Large (2) In the Upper Midwest, the presence of A. canescens is Wildlife Area [1] Jackson MN 1.1 Small (1) indicative of natural, undisturbed prairie plant communi- Sangl Slough [2] Jackson MN 0.5 Small (1) ties, based on its intolerance to soil disturbance and cattle Sangl Wildlife Area [3] Jackson MN 0.9 Small (2) grazing, and its Wdelity to upland prairie and savannah hab- Brouillette Property [4] Dickinson IA 0.3 Small (1) a itats (Swink and Wilhelm 1994), as well as its importance 170th Ave. Roadside [5] Dickinson IA 0.1 Small (1) a as a food resource for herbivorous and pollinating 200th St. Roadside [6] Dickinson IA 0.1 Small (1) (Lewis 1999). Mowing (Weaver 1954; McGregor et al. Floete Prairie [7] Dickinson IA 0.2 Small (1) 1986) or burning (Towne and Knapp 1996) stimulates Iowa Lakeside Labb [8] Dickinson IA 0.4 Small (1) growth of new herbaceous ramets, each of which can pro- Iowa Northwestern RRb [9] Dickinson IA 0.6 Small (3) duce one large, compound raceme. In the absence of Wre or Indian Hills Golf Course [10] Dickinson IA 0.2 Small (2) mowing, A. canescens becomes woody and can grow up to Milford Bike Trail [11] Dickinson IA 0.2 Small (2) 1 m tall with stems 1–2 cm in diameter, each of which can Letters and numbers in brackets are map references (see Fig. 1). Num- bear many small racemes (Weaver 1954). The sensitivity to ber of study plots at each site is given in parentheses following site management and negative eVects of habitat degradation on size: large (>40 ha) or small (< 2 ha) A. canescens make it an important plant for an analysis of a 170th Ave. Roadside was only sampled in 2002, was mowed in 2003, its reproductive response to habitat fragmentation. A vari- and was replaced by 200th St. Roadside in 2003 b ety of herbivores feed on A. canescens, but post-pollination Two Iowa Northwestern Railroad remnants (RR) sampled in 2002 did not Xower in 2003. One of these sites was replaced by a sampling fruit predation is less than 5% in this species (Lewis 1999). plot in a nearby remnant, the other by a remnant at Iowa Lakeside Lab

Study sites three at each preserve for a total of 21 plots. In 2003 we studied 18 of the original 21 plots that contained enough In the summers of 2002 and 2003 we studied A. canescens A. canescens racemes for bee sampling and added three in dry to mesic upland prairies in 11 sites consisting of one new plots (Table 1). or more small remnants (<2 ha) and three sites that are large state prairie preserves (45–80 ha; Table 1; Fig. 1). Bee sampling The study sites used here have been identiWed as remnants in previous Xoral studies in the region (Glass 1981; Lewis In each plot, during each of three sampling periods, we 1999) and they contain native plant species susceptible to chose 150 racemes bearing many open Xowers to sweep degradation. All sites are located in northwest Iowa and sample for bees. To reduce sampling error as a result of col- southwest Minnesota within a 20 km by 30 km area; the lector bias, only one of the authors (M. W. S.) sampled minimum distance between sites was 1 km. The sites with bees. Because diVerent species of bees are active at diVer- small remnants are located in railroad right-of-ways, ent times of day (Stone et al. 1999), bees were captured in wildlife management areas, privately managed areas, and each plot on the 150 Xowering racemes of A. canescens in roadsides. Landscapes up to 2 km surrounding two sites three sample periods, one in the morning (0800–1100 (Anderson Prairie and Cayler Prairie) and landscapes sur- hours), one midday (1100–1400 hours), and one late after- rounding 13 similar small remnants to ones used here have noon (1400–1700 hours). For any given site, sampling been characterized in detail with respect to variation in during the three daytime periods was on diVerent days landscape elements such as Xoral resources available to spread out through the 4- to 5-week Xowering period with Xower visitors (Davis et al. 2007). Generally, Xoral resources the order of periods sampled chosen randomly for diVerent available in the landscape surrounding these sites in sites. Due to bees’ sensitivity to weather conditions, sam- northwest Iowa are low. pling was conducted only if the temperature was greater In 2002 we established 1,000-m2 plots at study sites than 16°C, and the wind speed was less than 37 km/h, within the densest portions of populations of A. canescens threshold conditions for bee activity and collecting in to maximize pollinator activity within our plots. We estab- northwestern Iowa (M. Slagle and S. Hendrix, personal lished one to three plots at each small remnant and two or observation). Excluding handling time and walking 123 816 Oecologia (2009) 161:813–823

Fig. 1 Location of study sites in Dickinson County, Iowa and Jackson County, Minnesota. Filled square Large Iowa state prairie preserves (>40 ha), Wlled circle small private/public rem- nants (<1.5 ha), dark gray areas bodies of water, light gray areas municipalities. A Cayler Prairie, B Freda HaVner Kettle- hole, C Anderson Prairie, 1 Wildlife Area, 2 Sangl Slough, 3 Sangl Wildlife Area, 4 Brouil- lette Property, 5 170th Ave. Roadside, 6 200th St. Roadside, 7 Floete Prairie, 8 Iowa Lakeside Laboratory (North 40), 9 Iowa Northwestern Railroad rem- nants, 10 Indian Hills Golf Course, 11 Milford bike trail

between clumps, we searched racemes in each plot for oligolectic species that collect pollen from only a few spe- 20 min in each of the three sampling periods for a total of cies, but visit many more for nectar (Cane and Sipes 2006). 1 h per site per year. The number of bees captured during these three sampling periods was totaled to calculate bee Plant sampling abundance per plot per year. We identiWed collected bees to genus using Michener In 2002 and 2003 we estimated the abundance of racemes et al. (1994), to subgenus using Michener (2000), and to within plots as follows. We Wrst directly counted the num- species using appropriate monographs. Bees in the genus ber of A. canescens clumps per plot. We then counted the Lasioglossum, subgenera Dialictus and Evylaeus, could not number of racemes per clump in those clumps sampled for be identiWed to species because no reliable key for these bees 3 times during the Xowering season. We averaged our subgenera is available (R. McGinley, Illinois Natural three counts to estimate the mean number of racemes per History Survey, personal communication). Therefore, we clump and multiplied this value by the number of clumps to grouped these subgenera into eight morphospecies using estimate raceme abundance per plot. We estimated racemes Mitchell (1960) as a guide. Given the morphological diver- per site by counting the number of A. canescens clumps in sity of these subgenera, this number of species is a conser- regularly spaced quadrats and multiplying that by the num- vative estimate. ber of racemes per clump. We classiWed bee species as polylectic or oligolectic In 2002 and 2003, as A. canescens was beginning to based on previously published studies (e.g., Mitchell 1960, Xower, we identiWed and recorded the bee-visited Xowering 1962; Hurd 1979 and references therein) and by compari- forb species inside the plots to measure the overall Xoral son with extensive records (5,000 specimens) from ongoing species density per plot or plant species density, the number surveys of the whole community of bees on Xowers in of plant species in a given area. In 2003 we also counted the Iowa. Bees were netted from Xowers and the plant species number of Xowering ramets of each forb species occurring on which the bee was visiting when caught was noted. We within the plot to determine bee-visited Xoral diversity. We classiWed bee species as “oligolectic” according to the crite- identiWed Xowering forb species using McGregor et al. ria of Cane and Sipes (2006) who deWne them as species (1986) as our primary reference. We classiWed forbs as restricted to visiting several related genera throughout “bee-visited” if the plant species was visited by bees in a their geographic range despite a wider array or “menu” of survey of the community of bees visiting prairie forb possible pollen sources. Visitation data may “hide” some species in northwest Iowa (S. Hendrix and K. Gaddis, 123 Oecologia (2009) 161:813–823 817

University of Iowa, unpublished data) or in the surveys of between potential continuous predictors of fruit set, bee Robertson (1928) of the pollinator communities on forb diversity, and bee abundance. In all models we only used species in Illinois. predictors showing no collinearity (P > 0.10). Models con- tained a maximum of three to four variables. In regression Fruit set estimation analyses we also examined variance inXation factors of pre- dictors, using the conservative threshold value of >5 based To measure A. canescens’ reproductive success, we col- on regressions of each independent variable on all others to lected the closest mature raceme to each of 20 randomly identify excessive collinearity (Freund and Littell 1986). chosen points within each of the plots in early September Variance inXation factors were below threshold for all 2002 and 2003. We counted the number of fully developed, results reported. We represented each of the two sizes of undamaged fruits on each raceme. Then, we measured the remnants as dummy variables (small remnants = 1; large length of each branch of the raceme and totaled the branch remnants = 0) in order to be able to include them as a com- lengths for each raceme. To estimate the total number of ponent of the regression analysis (Daniel 1995). We used open Xowers on a raceme that could potentially produce a maximum R stepwise linear regression analysis (Draper fruit (one fruit/Xower), we used the following equation 1998) to examine relationships between percent fruit set, derived from undamaged Xowering A. canescens racemes bee diversity, and bee abundance and predictor variables. In randomly chosen at Cayler Prairie State Preserve (S. Hendrix, the multiple linear regression analyses, we excluded plots unpublished data): that were sampled for bees only 2 times over the Xowering Total Xowers = 10.645 + (21.89 £ total sum of all season (three plots in 2002, one plot in 2003) because they branch lengths in a raceme in cm) (R2 = 0.81, P = 0.0001, had signiWcantly lower bee abundance than plots sampled 3 n = 29). times (Mann–Whitney U-test: 2002, U3,18 =48.5, P <0.05; We then divided the total actual fruit count by the esti- 2003, U1,20 =20, P =0.05). mated total number of Xowers for each raceme (£ 100) to We chose best models based on values of Mallow’s t  obtain percent fruit set. We averaged the percent fruit set Cp p (the number of model parameters including 0) and of the 20 racemes within each plot to obtain an estimate of by examining Cp plots of all models (Draper 1998). For all the mean percent fruit set per raceme for each plot. Post- regressions, we conducted Cook’s D-tests and removed any fertilization seed predation was noted, but was rare. extreme outliers (D >4/n). All statistical analyses were performed using SAS version 8.2 (SAS Institute 2001). Statistical analysis

We used the log2 Shannon–Wiener index of diversity Results (Magurran 2004) to estimate bee diversity on A. canescens and Xoral diversity per plot. For all variables in sites with Fruit set multiple plots, we averaged the values to get one value for the site. We checked the normality of all continuous predic- The best linear regression models of population-level fruit tor and response variables with the Kolmogorov-Smirnov set in Amorpha canescens showed signiWcant relationships test and transformed two variables. In both years, the num- with bee community characteristics in both years (Table 2). ber of A. canescens racemes in a plot was log transformed In 2002 the best model included two signiWcant variables, and percentage of Xowers setting fruit was arcsin trans- bee diversity (partial r = 0.407, P = 0.035; Fig. 2a) and bee- formed. Abundance values are based on seasonal total num- visited plant species density (partial r = 0.335, P=0.012), ber of bees per plot, with the number averaged for the sites which were signiWcantly and positively related to fruit set. with multiple plots. Substitution of bee abundance for bee diversity in the mod- We used repeated measures ANOVAs using PROC els resulted in a single-variable best model with only total GLM (SAS Institute 2001) to compare the bee abundance bee abundance in it (Fig. 2b). Inclusion of remnant size and diversity in diVerent years and bee abundance at diVer- (large or small), measures of raceme abundance at the site ent times of day. Similarly, we used repeated measures (racemes/plot and total racemes), or bee functional diver- ANOVA to compare the average number of bee-visited sity did not contribute signiWcantly to any of the models (all plant species per site in diVerent years. For all year-to-year P-values >0.10). comparisons, we only used the sites for which we had data In 2003, the best model of fruit set explained much less for the 2-year period. We use F-values associated with variance than the best model for 2002 and only single-vari- Wilks’  to evaluate results of repeated measure analyses. able models were signiWcant (Table 2). There was a signiW- Prior to regression analysis we used Pearson product- cant relationship between fruit set and bee diversity moment correlation analysis to examine for collinearity (Fig. 2c) and a marginal relationship between fruit set and 123 818 Oecologia (2009) 161:813–823

Table 2 Summary of signiWcant results from maximum R multiple diversity associated with A. canescens included a signiW- regression analysis of percent fruit set of Amorpha canescens, species cant negative eVect of bee-visited plant species density at diversity and abundance of bees visiting A. canescens the site (Fig. 3) and a marginal positive eVect of site size Year Adjusted Partial (Table 2). The best model for bee abundance in 2003 included a signiWcant negative eVect of plant density of Model R2 Model P Predictor (sign) rP bee-visited species (Table 2). There was no signiWcant vari- a Percent fruit set ation between years in diversity (F1,9 =0.01, P >0.94) and 2002 0.742 <0.005 Bee diversity (+) 0.407 <0.035 abundance (F1,9 = 2.71, P > 0.10) of bees visiting A. canes- Plant density (+) 0.335 <0.012 cens. Similarly, bee-visited Xoral density also did not diVer W 2003 0.458 <0.02 Bee diversity (+) <0.02 signi cantly between years (F1,9 =1.12, P > 0.30). Bee Bee species diversityb abundance did not diVer at diVerent times of day in 2002

2003 0.578 <0.02 Plant density (¡) ¡0.634 <0.04 (F2, 9 =2.52, P > 0.15) or 2003 (F2,11 =1.20, P >0.30). Site size (+) 0.42 <0.07 Of the bee species we captured on A. canescens, we con- Bee abundancec sidered three species oligolectic: Andrena (Trachandrena) 2003 0.526 <0.01 Plant density (¡)<0.01quintilis Robertson (Andrenidae), Calliopsis andreniformis Smith () and Colletes robertsonii Dalla Torre For all tests: in 2002, n = 12; in 2003, n =13 (Colletidae). Andrena quintilis has a very strong preference a Predictor variables tested: bee abundance, bee diversity, size, abun- dance of racemes/plot, total site raceme abundance, and bee-visited for Amorpha pollen and nectar, although it will visit a num- plant species density (=“plant density”), and bee-visited plant species ber of other species for nectar and pollen (LaBerge 1973) diversity (2003) and Amorpha canescens is not obligately dependent on this b Predictor variables tested: size, raceme abundance/plot, total site species for pollination as A. quintilis does not cover the raceme abundance, bee-visited plant species density, bee-visited plant entire range of A. canescens (LaBerge 1973; McGregor species diversity (only in 2003) c et al. 1986). C. andreniformis is a “mesolectic” bee (Cane Predictor variables tested: raceme abundance/plot, total site raceme V abundance, bee-visited plant species density, and bee-visited plant and Sipes 2006) which visits a variety of genera in di erent species diversity (only in 2003) tribes within the subfamily in the Fabaceae (Dyer and Shinn 1978) as well as a number of other plant species for nectar and possibly pollen (Hurd 1979). bee abundance (Fig. 2d). Inclusion of remnant size, abun- Although Hurd (1979) considered C. robertsonii’s pollen dance of racemes, or functional diversity did not contribute sources to be unknown, it was reported to visit only Amor- signiWcantly to any of the models in 2003 (all P-values >0.22). pha, Dalea, and Monarda spp. (Stephen 1954). Cane (2006) considered the species an oligolege on Dalea purpu- Bee diversity and abundance rea; our collections of C. robertsonii from this and other studies support the status of oligolecty with 47 of 49 indi- We caught 49 species of bees on A. canescens at the 14 viduals from ten locations visiting A. canescens or Dalea study sites (Appendix). This is similar to the richness of bee spp. in sites with a broad menu of pollen sources. We con- species found by other researchers on A. canescens (Reed sidered the remaining 46 species polylectic (Cane and Sipes 1995). Honeybees were present at less than 30% of the sites 2006) because they were captured on or are known to visit in any one year and represented 3.0 and 6.4% of all bees species in many diVerent plant families. A. quintilis collected in 2002 and 2003, respectively. The species we accounted for most of the oligolectic bees collected, and captured on A. canescens in this study represented over it occurred at all sites, averaging 12.8 individuals per site. 40% of the bee species presently identiWed in the regional C. andreniformis occurred at only four of the study sites community on all Xowering plants (S. Hendrix and K. Gaddis, with abundances of one to three individuals. C. robertsonii University of Iowa, unpublished data). After two seasons occurred at seven sites, but only at abundances of one to and more than 40 total collection hours on A. canescens in two individuals per site. the region, 20 species (41% of all collected) were repre- sented by only one or two individuals in our collection. This level of rarity is not unusual for bee collections Discussion (Williams et al. 2001; Potts et al. 2003). Regression models of diversity and abundance of bees Our results indicate that of the possible predictors of repro- visiting A. canescens yielded mixed results. In 2002 none duction in A. canescens aVected by the fragmentation process of the models attempting to explain bee diversity and and examined in this study, pollinator services provided by abundance were signiWcant (all models adjusted r2 < 0.31, bees were the most consistently signiWcant (Table 2). The all P-values >0.32). In 2003, the best model for bee relationships between bee diversity and percent fruit set were 123 Oecologia (2009) 161:813–823 819

Fig. 2 Simple regression rela- tionships between arcsin-trans- formed mean percent fruit set (radians) of Amorpha canescens and a species diversity (Shan- non-Wiener log2 diversity) of bees visiting A. canescens in 2002 (y = 0.112x +0.287), b abundance of bees visiting A. canescens in 2002 (two outliers were removed; y = 0.009x +0.354), c species diversity (Shannon-Wiener log2 diversity) of bees visiting A. canescens in 2003 (y = 0.0586x + 0.347), and d abundance of bees visiting A. canescens in 2003 (y =0.003x + 0.408; one outlier removed). Open circle Large prairie preserves, Wlled circle small remnants

consistent across the 2 years of this study (Fig. 2a, c), while reductions in reproduction as a result of low pollinator the relationships between bee abundance and percent fruit diversity inXuence long-term demographic characteristics set were signiWcant for one year (Fig. 2b), and marginally is not known for A. canescens’ populations and has not signiWcant the other (Fig. 2d). Similarly, positive relation- been adequately examined empirically in other species ships between plant reproduction and pollinator abundance (Ashman et al. 2004). or diversity have been shown in small experimental plant- The lack of any signiWcant eVect of remnant size or ings (SteVan-Dewenter and Tscharntke 1999), small natural raceme abundance on fruit set in A. canescens in either year populations of an endangered species (Matsumura and of the study indicates that these characteristics of frag- Washitani 2000), and in Indonesian coVee plantations mented habitats are poor predictors of reproduction for where diversity was considered more important to pollina- A. canescens (Hendrix 1994; Lewis 1999). Our results are tion that abundance (Klein et al. 2003). in contrast to those of many studies that have shown nega- The signiWcant relationships between plant reproduction tive eVects of remnant size on reproduction in small popu- and bee species diversity may be due to diVerences in abil- lations of native plants (Jennersten 1988; Aizen and ity of species to successfully move pollen (Larsen et al. Feinsinger 1994; Ågren 1996; Morgan 1999; Brys et al. 2005) or because diVerent species occupy unique or nearly 2004), but are in agreement with others that have found no unique temporal and spatial visitation niche dimensions. If eVects or only intermittent eVects of size (Heschel and so, then redundancy in the prairie bee community is low Paige 1995; Alexandersson and Ågren 1996; Molano-Flores and functional equivalence in bee communities may be and Hendrix 1999; Gross et al. 2003). uncommon as it is in some other communities (Resetarits Our results showing no relationship between abundance and Chalcraft 2007; but see Bell 2007). Whether or not of A. canescens racemes and bee diversity or abundance in 123 820 Oecologia (2009) 161:813–823

initial eVects of habitat destruction rather than the later eVects of habitat fragmentation on bees (reviewed in Cane 2001), but because the destruction of the tall grass prairie in Iowa and Minnesota was completed 100 years ago, eVects of the initial destruction have likely subsided. The marginal positive eVect of site size (large preserves vs. small rem- nants) in 2003 on diversity of bees visiting A. canescens, indicates that the larger size may be attracting or conserv- ing more bee species, but this eVect was not observed in 2002, suggesting other factors may overwhelm eVects of site size diVerences in some years. The eVects of fragmen- tation on bee communities are likely to be complex given that responses of species may vary with respect to size, nesting habit, visitation patterns and other traits (Cane et al. 2006) and there can be variation in species composition from year to year. The negative relationship between both diversity and abundance of bees visiting A. canescens and Xoral species density in 2003 (Fig. 3) may result from other Xowering plants drawing some bee species away from A. canescens. However, we found no such relationship in 2002 and in fact there was a positive eVect of Xoral species density on repro- duction in 2002. We speculate that subtle diVerences between years in abundances and phenologies of both focal plant species and other co-occurring bee-visited plants may inXuence potential competition for pollinators. For exam- ple, the length of the Xowering period of A. canescens can vary from 40 to 60 days between years and the density of Xowering ramets at a site can vary up to threefold (Lewis 1999). In general, all prairie forbs can be expected to vary independently of each other from year to year in response to changing environmental conditions such as water avail- Fig. 3 The simple regression relationship between the mean density ability (Weaver 1954) and the rank order of most abundant of bee-visited plants per plot (1,000 m2) in 2003 and a species diversity forbs at Iowa prairie preserves and railway remnants can (log2 Shannon–Wiener index) of bees visiting A. canescens vary greatly from year to year (Hill 2005). (y = ¡0.128x – 3.975), and b mean abundance of bees visiting The high numbers (12.8 individuals/site) and widespread A. canescens (y = ¡2.60x + 55.68; one outlier removed). Open circle Large prairie preserves, Wlled circle small remnants occurrence of the oligolectic bee species Andrena (Trac- handrena) quintilis at all studied populations implies little discernable eVect of fragmentation on its population size either year and only a marginal relationship between bee and distribution. The high number of A. quintilis also abundance or diversity and Xoral species density one year implies an important role for it in reproduction by A. canes- (2003) are similar to a few other studies (Tepedino and cens, although no relationship was detected between abun- Stanton 1981; Becker et al. 1991; unpublished data dance of A. quintilis and fruit set of A. canescens. While the reviewed by Cane 2001), but are in contrast to other studies abundance of A. quintilis may not ensure future reproduc- showing that more diverse plant communities generally tion of A. canescens, its enduring presence and number at attract a greater abundance and/or species diversity of all populations in the area indicate that at least this oligolec- Xower visitors (Heithaus 1974; Powell and Powell 1987; tic species seems to be resilient for the time being to the Banaszak 1996; Reed 1995; Tscharntke et al. 1998; Potts eVects of fragmentation. The low abundance or limited et al. 2003) than less diverse ones. These studies examined occurrence of the other two oligolectic species is diYcult to visitors to all plant species rather than a focal plant species, interpret because in general oligolectic bees are temporally so it is unclear if all plant species in these diverse commu- and spatially unreliable at their host plants (Minckley and nities beneWt equally from increased pollinator diversity. Roulston 2006). One empirical study of oligolectic bees Additionally, in some cases, these studies represent the associated with Larrea tridentata showed that ground-nesting, 123 Oecologia (2009) 161:813–823 821 but not cavity-nesting, oligleges were negatively aVected Ashman T, Knight TM, Steets JA, Amarasekare P, Burd M, Campbell by remnant size (Cane et al. 2006), indicating a variable DR, Dudach MR, Johnston MO, Mazer SJ, Mitchell RJ, Morgan M, Wilson WG (2004) Pollen limitation of plant reproduction: response of specialist bees to fragmentation. In this study, ecological and evolutionary causes and consequences. Ecology all the oligolectic species are ground-nesters (Michener 85:2408–2421 2000). Alternatively, the low abundance of some oligolec- Banaszak J (1992) Strategy for conservation of wild bees in an agricul- tic species associated with A. canescens may merely reXect tural landscape. Agric Ecosyst Environ 40:179–192 Banaszak J (1996) Ecological bases of conservation of wild bees. In: their historically low numbers or the necessarily incomplete Matheson A, Buchmann SL, O’Toole C, Westrich P, Williams IH sampling represented by this short study. Monitoring of oli- (eds) The conservation of bees. Academic Press, London, pp 55– golectic bee populations should be continued in order to 62 increase understanding of these pollinators. Becker P, Moure JS, Peralta FJA (1991) More about euglossine bees in Amazonian forest fragments. Biotropica 23:586–591 In conclusion, our results indicate that at tall grass prai- Bell JJ (2007) Contrasting patterns of species and functional composi- rie remnants high diversity and abundance of bee visitors tion of coral reef sponge assemblages. Mar Ecol Prog Ser 339:73– can increase the reproduction of perennial forbs like 81 A. canescens even in relatively small populations located in Bock CE, Jones ZF, Bock JH (2008) The oasis eVect: response of birds to exurban development in a southwestern savanna. Ecol Appl fragmented landscapes. Even small remnants may function 18:1093–1106 as “resource oases” (Bock et al. 2008) for bees in land- Brys R, Jacquemyn H, Endels P, van Rossum F, Hermy M, Triest L, scapes often dominated by agriculture (Davis et al. 2007) De Bruyn L, Blust GDE (2004) Reduced reproductive success in which may account for the similar abundance and diversity small populations of the self-incompatible Primula vulgaris. J Ecol 92:5–14 of bees at small remnants and large preserves (Davis et al. Buchmann SL, Nabhan GP (1996) The forgotten pollinators. Island 2008). Furthermore, A. canescens is likely a very important Press, Washington, DC pollen and nectar “resource oasis” plant for bees because of Cane JH (2001) Habitat fragmentation and native bees: a premature its widespread presence in undisturbed prairies, long Xow- verdict? Conserv Ecol 5:3. [online] URL: http://www.ecology- X andsociety.org/vol5/iss1/art3/ ering period (Lewis 1999) and the numerous owers per Cane JH (2006) An evaluation of pollination mechanisms for purple raceme each of which produces an unusually large amount prairie-clover, Dalea purpurea (Fabaceae: ). Am Midl of pollen (mean = 16,000 grains) for a forb requiring a pol- Nat 156:193–197 linator (Cruden 1977). The high diversity and abundance of Cane JH, Sipes S (2006) Floral specialization by bees: analytical meth- ods and a revised lexicon for oligolecty. In: Waser NM, Ollerton bees visiting such small plant populations may be possible J (eds) Plant–pollinator interactions: from specialization to gener- because solitary bees can Xy relatively long distances alization. Chicago Press, Chicago, pp 99–122 between nest sites and Xoral resources (Greenleaf et al. Cane JH, Minckley RL, Kervin LJ, Roulston TH, Williams NM (2006) 2007). However, such small sites are unlikely to be able Complex responses within a desert bee guild (: Ap- iformes) to urban habitat fragmentation. Ecol Appl 16:632–644 to sustain bees without resources from the surrounding Cruden RW (1977) Pollen-ovule ratios: a conservative indicator of landscape (Kremen et al. 2004; Greenleaf and Kremen breeding systems in Xowering plants. Evolution 31:32–46 2006), which have been greatly reduced as a result of Daniel WW (1995) Biostatistics: a foundation for analysis in the health fragmentation. sciences, 6th edn. Wiley, New York Davis JD, Debinski DM, Danielson BJ (2007) Local and landscape eVects on the butterXy community in fragmented Midwest USA Acknowledgments We thank H. Snowgren, K. Gaddis, A. Hill, W prairie habitats. Landsc Ecol 22:1341–1354 M. Wendel, and C. Slagle for eld and laboratory assistance, Davis JD, Hendrix SD, Debinski DM, Hemsley CJ (2008) ButterXy, K. Cowles for statistical assistance, the Snow Entomological Museum, bee, and forb community composition and cross-taxon incongru- C. Gienapp, and C. Michener for bee identiWcation assistance, K. Copp W ence in tallgras prairie fragments. J Insect Conserv 12:69–79 for gure preparation, and the Avis Cone Foundation, Center for Draper NR (1998) Applied regression analysis. Wiley, New York Global and Regional Environmental Research, the American Museum andrenifor- W Dyer JG, Shinn AF (1978) Pollen collected by of Natural History and the Explorer’s Club for nancial assistance. mis Smith in (Hymenoptera: Andrenidae). J Kans These studies comply with current laws in the country in which they Entomol Soc 51:787–795 were conducted. Freund R, Littell RC (1986) SAS system for regression. SAS Institute, Cary Gamfeldt L, Hillebrand H, Jonsson PR (2008) Multiple functions increase the importance of biodiversity for overall ecosystem References functioning. Ecology 89:1223–1231 Glass WD (1981) The importance of refuge size in preserving species Ågren J (1996) Population size, pollinator limitation, and seed set in the of prairie legumes, goldenrods, and milkweeds. M.Sc. thesis. self-incompatible herb Lythrum salicaria. Ecology 77:1779–1790 University of Illinois at Chicago Circle, Chicago Aizen MA, Feinsinger P (1994) Habitat fragmentation, native insect Goulson D, Darvill B (2004) Niche overlap and diet breadth in pollinators, and feral honeybees in Argentine chaco serrano. Ecol bumblebees; are rare species more specialized in their choice of Appl 4:378–392 Xowers? Apidologie 35:55–63 Alexandersson R, Ågren J (1996) Population size, pollinator visitation Greenleaf SS, Kremen C (2006) Wild bee species increase tomato and fruit production in the deceptive orchid Calypso bulbosa. production and respond diVerently to surrounding land use in Oecologia 107:533–540 northern California. Biol Conserv 133:81–87 123 822 Oecologia (2009) 161:813–823

Greenleaf SS, Williams NM, Winfree R, Kremen C (2007) Bee foraging Michener CD, McGinley RJ, Danforth BN (1994) The bee genera of ranges and their relationship to body size. Oecologia 153:589–596 North and Central America (Hymenoptera: Apoidea). Smithso- Gross CL, Bartier FV, Mulligan DR (2003) Floral structure, breeding nian Institution Press, Washington, DC system and fruit-set in the threatened sub-shrub Tetratheca juncea Minckley RL, Roulston TH (2006) Incidental mutualisms and pollen Smith (Tremandraceae). Ann Bot 92:771–777 specialization among bees. Oligolecty. In: Waser NM, Ollerton J Hayes J (2007) Florida workshop discusses colony collapse disorder. (eds) Plant–pollinator interactions: from specialization to gener- Am Bee J 147:295–303 alization. Chicago Press, Chicago, pp 69–98 Heithaus (1974) The role of plant–pollinator interactions in determin- Mitchell TB (1960) Bees of the eastern , vol I. The North ing community structure. Ann Mo Bot Gard 61:675–691 Carolina Agricultural Experimental Station, Raleigh Hendrix SD (1994) EVects of population size on fertilization, seed pro- Mitchell TB (1962) Bees of the eastern United States, vol II. The North duction, and seed predation in two prairie legumes. In: Wickett Carolina Agricultural Experiment Station, Raleigh RG, Lewis PD, WoodliVe A, Pratt P (eds) Proceedings of the 13th Molano-Flores B, Hendrix SD (1999) The eVects of population size North American Prairie Conference, Department of Parks and and density on the reproductive output of Anemone canadensis L. Recreation, Windsor, pp 115–121 (Ranunculaceae). Int J Plant Sci 160:759–766 Hendrix SD, Kyhl JF (2000) Population size and reproduction in Phlox Morgan JW (1999) EVects of population size on seed production and pilosa. Conserv Biol 14:304–313 germinability in an endangered, fragmented grassland plant. Heschel MS, Paige KN (1995) Inbreeding depression, environmental Conserv Biol 13:266–273 stress, and population size variation in scarlet gilia (Ipomopsis Norton C (2007) Colony collapse disorder investigated by US bee aggregata). Conserv Biol 9:126–133 researchers. Am Bee J 147:199–202 Hill AF (2005) Diversity, abundance, and similarity of Xoral resources Packer L, Zayed A, Grixti JC, Ruz L, Owen RE, Vivallo F, Toro H for pollinators in tallgrass prairie fragments. M.Sc. thesis. Univer- (2005) Conservation genetics of potentially endangered mutual- sity of Iowa, Iowa City isms: reduced levels of genetic variation in specialist versus Hurd PD Jr (1979) Superfamily Apoidea. In: Krombein KV, Hurd PDJ, generalist bees. Conserv Biol 19:195–202 Smith DR, Burks BD (eds) Catalog of Hymenoptera in North Parrish JAD, Bazzaz FA (1979) DiVerence in pollination niche rela- America north of Mexico, 1st edn. Smithsonian Institute Press, tionships in early and late successional communities. Ecology Washington, DC 60:597–610 Institute SAS (2001) SAS software: changes and enhancements Potts SG, Vulliamy B, Dafni A, Ne’eman G, Willmer P (2003) Linking through release 8.2. SAS Institute, Cary bees and Xowers: how do Xoral communities structure pollinator Jennersten O (1988) Pollination in Dianthus deltoides (Caryophylla- communities? Ecology 84:2628–2642 ceae)—eVects of habitat fragmentation on visitation and seed set. Powell AH, Powell GVN (1987) Population dynamics of male Conserv Biol 2:359–366 euglossine bees in Amazonian forest fragments. Biotropica Klein AM, SteVan-Dewenter I, Tscharntke T (2003) Fruit set of high- 19:176–179 land coVee increases with the diversity of pollinating bees. Proc R Reed C (1995) Species richness of insects on prairie Xowers in south- Soc Lond B Biol Sci 270:955–961 eastern Minnesota. In: Hartnett DC (ed) Proceedings of the 14th Klein AM, Cunningham SA, Bos M, SteVan-Dewenter I (2008) North America Prairie Conference: Prairie Biodiversity. Kansas Advances in pollination ecology from tropical plantation crops. State University, Manhattan, pp 103–115 Ecology 89:935–943 Resetarits WJ, Chalcraft DR (2007) Functional diversity within a mor- Kremen C, Williams NM, Bugg RL, Fay JP, Thorp RW (2004) The phologically conservative genus of predators: implications for area requirements of an ecosystem service: crop pollination by functional equivalence and redundancy in ecological communi- native bee communities in California. Ecol Lett 7:1109–1119 ties. Funct Ecol 21:793–804 Kremen C, Williams NM, Aizen MA, Gemmill-Herren B, LeBuhn G, Richards CM, Church S, McCauley DE (1999) The inXuence of popu- Minckley R, Packer L, Potts SG, Roulston T, SteVan-Dewenter I, lation size and isolation on gene Xow by pollen in Silene alba. Vazquez DP, Winfree R, Adams L, Crone EE, Greenleaf SS, Keitt Evolution 53:63–73 TH, Klein AM, Regetz J, Ricketts TH (2007) Pollination and other Robertson C (1928) Flowers and insects. Lists of visitors to four hun- ecosystem services produced by mobile organisms: a conceptual dred and Wfty-three Xowers. Science Press, Lancaster framework for the eVects of land-use change. Ecol Lett 10:299–314 Roubik DW (2001) Ups and downs of pollinator populations: when is LaBerge W (1973) A revision of the bees of the genus Andrena of the there a decline? Conserv Ecol 5:2 [online] URL: http://www. western hemisphere. Part VI. Subgenus Trachandrena. Trans Am ecologyandsociety.org/vol5/iss1/art2/ Entomol Soc 99:235–371 SteVan-Dewenter I, Tscharntke T (1999) EVects of habitat isolation on Larsen TH, Williams NM, Kremen C (2005) Extinction order and pollinator communities and seed set. Oecologia 121:432–440 altered community structure rapidly disrupt ecosystem function- Steinauer EM, Collins SL (1996) Prairie ecology: the tallgrass prairie. ing. Ecol Lett 8:538–547 In: Samson FB, Knopf FL (eds) Prairie conservation: preserving Leach MK, Givnish TJ (1996) Ecological determinants of species loss North America’s most endangered ecosystem. Island Press, in remnant prairies. Science 273:1555–1558 Washington, DC, pp 39–52 Lewis CK (1999) The eVects of habitat fragmentation on Amorpha Stephen WP (1954) A revision of the bee genus Colletes in America canescens, a prairie forb, and its associated herbivores. M.Sc. thesis. north of Mexico. Univ Kans Bull 36:149–527 University of Iowa, Iowa City Stone GN, Gilbert F, Willmer P, Potts S, Semida F, Zalat S (1999) Magurran AE (2004) Measuring biological diversity. Blackwell Scien- Windows of opportunity and the temporal structuring of foraging tiWc, Malden activity in a desert solitary bee. Ecol Entomol 24:208–221 Matsumura C, Washitani I (2000) EVects of population size and polli- Swink F, Wilhelm G (1994) Plants of the Chicago region. Indiana nator limitation on seed-set of Primula sieboldii populations in a Academy of Science, Indianapolis fragmented landscape. Ecol Res 15:307–322 Tepedino V, Stanton N (1981) Diversity and competition in bee–plant McGregor RL, Barkley TM, Brooks RE, SchoWeld EK (1986) Flora of communities on short-grass prairie. Oikos 36:35–44 the great plains. University of Kansas Press, Lawrence Towne EG, Knapp AK (1996) Biomass and density responses in tall- Michener CD (2000) The bees of the world. Johns Hopkins University grass prairie legumes to annual Wre and topographic position. Am Press, Baltimore J Bot 83:175–179 123 Oecologia (2009) 161:813–823 823

Tscharntke T, Gathmann A, SteVan-Dewenter I (1998) Bioindication Weaver J (1954) North American prairie. Johnsen, Lincoln using trap-nesting bees and wasps and their natural enemies: com- Weaver JE, Fitzpatrick TJ (1934) The prairie. Ecol Monogr 4:109–295 munity structure and interactions. J Appl Ecol 35:708–719 Williams NM, Minckley RL, Silveira FA (2001) Variation in native Waser NM, Chittka L, Price MV, Williams NM, Ollerton J (1996) bee faunas and its implications for detecting community changes. Generalization in pollination systems, and why it matters. Ecology Conserv Ecol 5:7. [online] URL: http://www.ecologyandsociety. 77:1043–1060 org/vol5/iss1/art7/

123