City size, regional landscape context, and local resource characteristics influe nce patterns of bee occurrence: A case

study from northwestern Costa Rica by Victoria A. Wojcik Pollinator Partnership, 423 Washington Street, 5th floor, San Francisco, CA 9411 1 University of California, Berkeley Department of Environmental Science, Policy, & Management, 132 Mulford Hall, Berkeley, CA 94620 E: victoriawojcik@hotmail.com; vw@pollinator.org T: 510-697-3890/415-362-1137 F: 415-362-3070 Abstract Understanding the ecological patterns of bees in cities can aid in their conserv ation and management. This is especially important in regions experiencing rapid anthropogenic change that threatens the survival of these important pollinators and the ecological services that they provide. To study occurrence trends at th e community level, bees visiting a common and abundant floral resource species, Tecoma stans, were sampled across three urban landscapes located in the tropical dry forest of northwestern Costa Rica. Frequency-of-visitation counts that meas ured both bee visitor abundance and bee taxon richness were used to assess forag ing variation and individual resource constancy, and to evaluate the effects of plant resource characteristics (floral abundance, resource size, and vertical st ructure), city size, and regional landscape context on bee occurrence. Of the re source characteristics studied, only the total number of flowers at a food resou rce had a significant impact on bee occurrence, with resources that possessed mo re flowers attracting more bees and more bee types. The taxon richness observed at a resource increased significantly as city size increased, but there was no a ssociation between city size and bee visitor abundance. Cities located within si milar regional landscapes had similar bee community composition, indicating that the surrounding landscape influenced species presence and absence patterns. The patterns of occurrence recorded in this study can be seen as a positive result for bee conservation and management in urban landscapes; the correlation of high bee visitation with abundant flowers suggests that efforts to increase resource availability should be successful in producing a corresponding increase in bee presence. Keywords: bees, hymenoptera, pollinators, foraging, resource characteristics, re gional landscape, tropical dry forest, Costa Rica, Tecoma stans Introduction: Bees are essential ecosystem service providers that are in large part responsibl e for maintaining terrestrial diversity and productivity. The association of bee s with flowering plants structures and sustains ecosystems (Kevan 1999; Kevan 20 01; Costanza et al. 1997) from which humans and other species derive direct and indirect benefits (Buchmann and Nabhan 1996; Daily 1997; Kevan and Wojcik 2007). Declines in populations of bees have been noted in recent years and are driven in part by land-use intensification, such as the conversion of wildlands into ag ricultural and urban landscapes (NRC 2007; Kremen et al. 2007; Winfree et al. 20 07; Winfree et al. 2009). Land-use intensification often diminishes or removes t he nesting and foraging habitats of native bees (Kremen et al. 2002). The tropic al dry forest region of Costa Rica is an example of an area of rapid urbanizatio n where declines in bee numbers have been noted at the interface of urban develo

pment (Frankie et al. 2009). Declines in the abundance and diversity of bees are paralleled by declines in pollination services (Kevan and Phillips 2001). Conti nued declines in pollination could have serious ecological and economic impacts, as one third of the food we eat is dependent on pollinators (Buchmann and Nabha n 1996), and the majority of flowering plants require pollinators for reproducti on (Kevan 1999). Early twentieth-century taxonomic surveys verify that bees have been present in urban landscapes for some time (Robertson 1929; Graenicher 1930). More recently, rich communities of bees have been documented in cities in North America (Frank ie et al. 2005b; Tommasi et al. 2004; Fetridge et at. 2008; Matteson et al. 2008 ), Central America (Frankie et al. 2009a), South America (Matteson et al. 2008; Loyola and Martins 2006; Nates-Parra 2006), Europe (Banaszak 1982; Berezin 1995; Saure 1996), and Asia (Sakagami and Fukuda 1973; Hisamatsu and Yamane 2006). Al though the presence of bee communities within cities does not suggest that urban landscapes can entirely compensate for lost habitat, the colonization of urban green space by bees might provide a buffer for some of the effects of wildland d egradation on pollination services. This observation has led scientists to inves tigate the conservation and management potential that human landscapes hold for ecologically and economically important species. Understanding why a species occurs at a site and what factors are important to t his occurrence can inform conservation planning and management. Associations bet ween bees and flowering plants are thought to be distinct and predictableseasonal ity, host-specificity, and daily flight times are patterned (Linsley 1958; Ginsb erg 1983; Wojcik et al. 2008). Resource usage and exploitation are often optimiz ed with the heavy metabolic demands of flight (Heinrich 1979). In tropical syste ms, floral resources occur at many strata with corresponding vertical stratifica tion, or structural niche partitioning, in different bee species (Roubik 1993). Bees also display site-specific preference patterns in their foraging behaviors, and not all resources within a landscape are visited equally. The study of bee ecology in urban landscapes has focused mostly on identifying species presence a nd absence, describing faunas, or outlining host plant interactions. More detail ed studies that outline local occurrence patterns are needed to develop a better understanding of bee foraging strategies and behaviors in cities. Click image to enlarge Figure 1: Examples of Tecoma stans plants growing in tropical urban landscapes a nd displaying a range of structural morphology. Clockwise from top left: (a) shr ub, (b) hedge, (c) tree, and (d) a close-up of the inflorescence. Tecoma stans Kunth (Bignoniaceae) is a common, mass-flowering, woody perennial f ound across Costa Rica and much of the tropics. In urban environments, it is pro minent in home gardens and city squares and along city streets. It has been used extensively as a decorative ornamental because of its year-round production of striking, yellow, trumpet-shaped flowers. Previous observations indicated that T . stans attracts a wide range of bee species from many genera, diverse guilds, a nd many foraging strata, representing over 10% of the local wildland bee fauna ( Wojcik 2009). In the three urban landscapes under study here, T. stans is ubiqui tous and presents substantial variability in plant form and structure, particula rly in urban landscapes where human management, such as pruning, has resulted in individuals that range from low-growing shrubs through full hedges to tall tree s (Figure 1). Its ubiquitous distribution, structural variability, and observed high attractiveness to bees make T. stans a good case-study species for investig ations of urban bee foraging preferences in multiple tropical cities. The goal of this study was to determine what makes a food resource located withi n a city attractive to local bee species. Are the structural and morphological p

atterns documented in wildland systems conserved in urban systems? Are all citie s similar, and do bees choose the same types of resources in any urban landscape ? Furthermore, given the diversity of bee species within urban systems, are all of these bees looking for, or responding to, the same factors? To answer these q uestions, a forager visitation survey was conducted in the cities of Bagaces, Caa s, and Liberia, Costa Rica, focusing on the community of bees that visited T. st ans resources. The recorded patterns of bee visitation were examined with respec t to resource characteristics (floral abundance, height, and plant form), city s ize, general local landscape characteristics, and regional landscape context to determine if any other patterns became evident. Gaining a better understanding o f how bee foraging strategies and behaviors manifest in cities will add valuable information to the new and growing discipline of urban bee ecology. Materials and Methods Site description The cities of Liberia (1037'47.19"N, 8526'17.75"W), Bagaces (1031'34.29"N, 8515'17.7 3"W), and Caas (1025'35.95"N, 8505'28.02"W) have developed along Central America Hi ghway 1 in the Guanacaste province of Costa Rica. Each city is about 25 kilomete rs from its neighbor, separated by remnant tropical dry forest that is fragmente d by ranching and agriculture. Urbanization and growth in this region are above the national average, with the population doubling in the past decade (INEC 2000 ). These three cities have experienced corresponding urban and peri-urban growth . The most recent national census puts the population of Liberia at 39,242 (34,4 69 urban; 4,773 rural), Bagaces at 9,261 (3,645 urban; 5,616 rural), and Caas at 18,798 (16,512 urban; 2,286 rural) (INEC 2000). The area currently covered by Li beria is approximately 8 square kilometers; Bagaces, the smallest city, covers a pproximately 1 square kilometer; and Caas covers approximately 3 square kilometer s (urban land coverage estimated in ArcMap 9.3, ESRI Inc. 2009). Demographic pro jections point to significant growth in this region; by 2015, Liberia is expecte d to have nearly 45,000 inhabitants, Caas will grow to over 38,000 inhabitants, a nd Bagaces will double to over 22,000 inhabitants (INEC 2000). Bagaces and Liber ia lie in close proximity to the biological reserve network of Lomas Barbudal an d Palo Verde and are surrounded by land that is used primarily for cattle grazin g. Caas resides within an intensive crop-production region. Building density and the size of individual lots is smaller in both Liberia and Bagaces than in Caas, but Caas has more public green space and parks. Bagaces and Liberia are bisected by many riparian zones, but there is only one riparian corridor in Caas. Tecoma stans (Bignoniaceae) is common along city streets and on private land in all three cities, as well as throughout the wildlands in this region. The native range of this species in the Americas extends from the south-central and southe astern United States (Arizona to Florida) to the Antilles and down into Argentin a (Hammel 2005; Zuchowski 2007). In Costa Rica, T. stans is most often found on Pacific slopes in Guanacaste province and into the central mountainous region, i ncluding the greater San Jos area of Puntanenas province (Hammel 2005). Individua l plants can reach a height of 10 meters (Zuchowski 2007) but are commonly short er in urban landscapes (Figure 1). Resource Classification The location of each plant in the study was recorded using a hand-held Garmin eT rex Vista C GPS unit. A digital photograph was taken at each site to serve as a r eference and to assist with resource height assessment. Height was estimated to the nearest 25 centimeters using either a meter measure or photographic scaling with the author (measuring 158 cm) as a reference. The T. stans individuals in t his study were then grouped into three height classes: small (< 1 meter), medium (13 meters), and large (> 3 meters). Plant form was categorized as shrub, hedge, or tree as follows: shrubs were low-growing (generally shorter than 1 meter) in

dividuals that lacked a clear dominant axis; hedges ranged in height and were pr edominantly characterized by having top-to-bottom foliage; trees had a distinct dominant axis (trunk) and a crown of foliage and flowers. The number of flowers that an individual T. stans possessed was estimated using a stratified subsample of the inflorescences in the canopy. The canopy was fract ioned into a manageable sector based on the overall resource size (either 2, 4, 8, or 10 sectors). The number of fractioned sectors was then used as an expansio n factor (k) to calculate the estimated number of inflorescences (i) and the sub sequent total flowers per resource (total flowers). The total number of inflores cences in one randomly selected sector was counted (i). A random subsample of th e enumerated inflorescences was then selected, and the total number of flowers c ontained in each inflorescence was counted and averaged over the total sample (). The average number of flowers per inflorescence was then multiplied by the numb er of inflorescences and by the expansion factor using the following equation: t otal flowers = x i x k. In the case of some individual plants that possessed ver y few flowers, the total number of flowers was enumerated using a census. T. sta ns individuals were then classified into three categories of floral resource abu ndance: low (< 100 flowers per individual), medium (100300 flowers per individual ), and high (>300 flowers per individual). Bee visitation assessment Between July 2007 and March 2009, 2,088 visitation counts documenting the richne ss and abundance of foraging bees were conducted at 120 unique T. stans individu als across the three landscapes. A standardized 1 x 1-meter square observation f rame was visually projected onto an easily observable and unobstructed area of t he plant (floral resource). The visits of bees to all of the flowers within this visual frame were counted for a period of three minutes. To achieve a represent ative sample of overall bee visitation, counts were taken approximately every 45 minutes during the daylight hours of 6 a.m. to 6 p.m. Efforts were made to rand omize the observation frames over repeated visits to each resource. A visitation was recorded only if a bee entered the corolla of the flower. Generalized flight patterns and overall gestalt (morphology, coloration, etc.) w ere used to identify bee visitors "on the wing". Eleven distinguishable taxonomi c categories of visitors were observed: African honey bees (Apis mellifera scute llata, abbreviated AHB), stingless bees (Trigona spp., abbreviated STB), Centris (Centris spp. and other resin bees), Eulema (Eulema spp.), Epicharis (Epicharis spp.), Euglossa (Euglossa spp.), Halictid (members of the sweat bee family Hali ctidae), Mesoplia (Mesoplia spp.), Melitoma (Melitoma spp.), Xylocopa (Xylocopa spp.), and small bees (when a specific identification could not be made). Abunda nce was recorded as the total number of individual bees that were observed per c ount. Taxon richness was recorded as the total number of different distinguishab le taxonomic categories observed per count. Representative samples of foraging bees were collected with aerial netting for m ore refined taxonomic identification. The bee collections were labeled and store d in standard solid wood entomological collection boxes. Bee identification was assisted by Laurence Packer and Jason Gibbs of York University, Toronto, Canada, and Ricardo Ayala of Universidad Nacional Autnoma de Mxico (UNAM). The majority o f the collection resides at the University of California, Berkeley Essig Museum of Entomology, with some parts at the Pollinator Partnership offices in San Fran cisco, California. The bee-plant interactions described here are also catalogued with the Pollinator Thematic Network and can be found by searching the bee spec ies listed in the Appendix. African honey bees (AHB), stingless bees (STB), members of the family Halictidae (Halictids), and bees in the genus Centris (Centris) were the dominant groups o bserved in this study. Statistical analysis that aimed to describe trends in fun

ctional groups focused on these four categories. Statistical Analysis The observed variability in abundance and taxon richness seen at Tecoma stans re sources was verified using repeated measures ANOVA (nresource = 42, nrepeats = 3 , = 0.05). Visit tion d t were n lyzed collectively s well s by city to inve stig te site-specific v ri bility. Multiv ri te ANOVA w s lso used to est blish if t xon richness, visitor bund nce, nd individu l bee groups differed betwee n the three cities studied. A Bonferroni djusted of 0.016 (0.05/3) w s used in this comp rison to djust for Type 1 Error ssoci ted with multiple tests. The m in nd inter ction effects of resource ch r cteristics (height, life-form, nd flor l bund nce) on bee bund nce, richness, nd the occurrence r tes of t he four domin nt t xon groups (Centris, H lictid, AHB, nd STB) were ex mined us ing multiv ri te ANOVA, with Bonferroni djusted of 0.016 (0.05/3) to djust f or Type 1 Error. All of the multiv ri te tests were further n lyzed with Tukey' s HSD to determine the direction nd m gnitude of the observed differences. To d etermine if there were ny ssoci tions between visit tion r tes nd city size o r the surrounding l ndsc pe, the bund nce, richness, nd visit tion r tes of th e individu l bee groups were comp red using ANOVA, =0.05, followed by Tukey's HSD for multiple comp risons. All st tistic l n lysis w s run in SPSS 16 (Rel. 16. 0.0, 2007 Chic go: SPSS Inc.). Results

V ri bility in individu l Tecom st ns resources St tistic l tests verified the observed v ri bility in bee ttr ction to individ u l T. st ns pl nts to be signific nt. Between-resource bee v ri tion r tes were signific ntly different for both bund nce (F1,441 = 0.005744, p = 0.000) nd t xon richness (F1,441 = 0.003596, p = 0.000). Me sures of within-resource bee vi sitor bund nce nd t xon richness exhibited contr sting trends. Bee bund nce d id not v ry signific ntly from me sure to me sure on the s me resource (ANOVA, F 2,882 = 0.517, p = 0.596); however, the number of different bee t x recorded pe r count did (repe ted me sures ANOVA, F2,882 = 2.329, p = 0.098). In ddition, t here w s signific nt inter ction effect between me sures of bee t xon richness nd visitor bund nce nd the individu l resource s mpled (p < 0.000), which su ggests th t other v ri bles might be structuring p tterns of bee visit tion to T . st ns in urb n l ndsc pes. Food resource ch r cteristics nd bee for ging Click im ge to enl rge Figure 2: The me n bund nce nd richness of bees ttr cted per count to T. st n

The bees observed nd collected from Tecom st ns represent 27 species in 20 gen er nd 2 f milies. The species were diverse in their feeding styles, nesting h bit ts, nd degrees of soci liz tion (Appendix). Most visitors to T. st ns in th e urb n l ndsc pe were solit ry, ground-nesting species; however, c vity-, twig, nd wood-nesters, s well s eusoci l nd commun l species, were present. Thre e groups of n tive bees ccounted for 90.36% of the tot l visitors recorded: Cen tris, STB, nd H lictid. Centris w s the most bund nt t 52.10%, followed by ST B t 19.43%, nd H lictid t 18.86%. Nonn tive AHB (Apis mellifer scutell t ) ccounted for 5.68% of the tot l records. The rem ining seven groups ccounted fo r less then 2% e ch of the tot l visitors in the entire s mple: Eulem , 1.12%; X ylocop , 0.89%; Eugloss , 0.69%; Epich ris, 0.44%; Mesopli , 0.19%; sm ll bee, 0 .16%; nd Melitom 0.15%.

Bee bund nce nd richness were signific ntly influenced only by the number of f lowers th t resource h d (F2,16 = 4.014, p = 0.000; F2,16 = 4.836, p = 0.000, respectively), nd not by its height or pl nt form (T ble 1). To lesser extent , the inter ction between tot l flowers nd pl nt form influenced the bund nce of bees visiting resource (F4,16 = 2.063; p < 0.091), but not the tot l number of bee types. Figure 2 displ ys the gener l trends seen in the bund nce nd ri chness of bees observed per count s function of resource ch r cteristics. The three flor l cl sses re signific ntly different from e ch other; in order of i ncre sing flor l bund nce, e ch c tegory ttr cted signific ntly more bees nd more bee types (Tukey's HDS; p >0.05) th n the one before. There w s no observed v ri bility in visitor bund nce nd ttr cted t xon richness b sed on the pl n t form. Although medium nd t ll resources did ttr ct more bees nd more bee ty pes th n shorter resources, this difference w s not signific nt.

City size nd ch r cteristics of the surrounding region l l ndsc pe E ch city h d unique bee f un nd specific p tterns of for ger occurrence. Th e me n occurrences of 9 of the 11 bee groups s mpled v ried signific ntly betwee n the 3 cities (T ble 3). Only Euglossine bees nd sm ll bees did not show ny s ignific nt v ri tion (F2,26 = 1.825; p = 0.161 nd F2,26 = 2.044; p = 0.130, res pectively). The size of the city h d signific nt imp ct on the number of bee t x th t were observed per count. Me n t xon richness incre sed signific ntly in the order B g ces < C s < Liberi (Tukey's HSD, B g ces: me n = 1.150, 95% CI [ 1.070, 1.230]; C s: me n = 1.368, 95% CI [1.294, 1.442]; Liberi : me n = 1.479, 95% CI [1.413, 1.545]); which is consistent with n incre se in size from the sm llest to the l rgest urb n l ndsc pe (Figure 3). City size did not h ve line r correl tion with bund nce p tterns. Click im ges to enl rge Figure 3: The me n number of different bee t x observed per three-minute count cross ll counts t ken in the three study cities, B g ces, C s, nd Liberi . Th e cities re rr nged on the x- xis in order of incre sing size. In e ch c se, t here re signific ntly more bee types per count s city size incre ses (Tukey's HSD, B g ces: me n = 1.150, 95% CI [1.070, 1.230]; C s: me n = 1.368, 95% CI [1. 294, 1.442]; Liberi : me n = 1.479, 95% CI [1.413, 1.545]). Error b rs represent the 95% confidence interv ls for e ch me sure. Figure 4: The represent tive proportions of the domin nt bee groups (AHB, Centri s, STB, nd H lictids) recorded on T. st ns resources in the cities of B g ces,

When the four domin nt bee groups were ex mined for their individu l responses t o resource ch r cteristics, more specific p tterns were seen (T ble 2). The for ging r tes of Centris nd STB were not signific ntly imp cted by ny of the reso urce ch r cteristics or their inter ctions. The for ging r tes of H lictids were imp cted by the pl nt form nd the tot l number of flowers possessed by resou rce (F2,16 = 5.477; p < 0.005 nd F2,16 = 4.764; p < 0.010, respectively), s we ll s by the inter ction between the two (F2,16 = 5.094; p = 0.001). AHB were si gnific ntly influenced by the tot l number of flowers resource h d (F2,16 = 8. 540; p = 0.000), nd by the inter ction between pl nt form, height, nd tot l fl owers (F3,16 = 7.472; p = 0.007). The community-level for ging p tterns re like ly driven by the responses of H lictids nd AHB to individu l resource v ri bili ty.

s s function of three different resource ch r cteristics: height (short = < 1 meter; medium = 13 meters; t ll = >3 meters), pl nt form (shrub, hedge, tree), nd flor l bund nce (low = < 100 flowers; medium = 100300 flowers; high = >300 fl owers).

C s, nd Liberi . "Minor Groups" represents those bees th t ccounted for < 2% o f the tot l bund nce found t e ch site, nd in e ch c se included: Xylocop , E pich ris, Eugloss , Eulem , Mesopli , Melict , nd sm ll bees. The domin nt regi on l l nd use is indic ted in br ckets for e ch city. There w s lso rel tionship between t xon domin nce nd the gre ter l ndsc pe context. Figure 4 displ ys the community composition of bees observed in e ch ci ty over the three-ye r study period. B g ces nd Liberi re situ ted within c t tle r nching l nds nd ne rer to conserv tion l ndsc pes; they h ve simil r prop ortions of Centris, H lictids, nd STB. C s is situ ted within more gricultur l l ndsc pe nd h d higher numbers of Centris nd STB, but lower occurrences of H lictids. Discussion The tropic l dry forest of Cost Ric is known for its bee diversity. More th n 250 n tive bee species re found in the wildl nds of this region (Fr nkie et l. 1983). M ss-flowering ngiosperms common to the s v nn nd mesic h bit ts prov ide ide l for ge th t supports the bee community (Fr nkie et l. 2004), but r pi d urb n growth is tr nsforming this ecosystem. Tecom st ns resources loc ted in the three cities studied here ttr cted subs mple of the wildl nd bee communi ty27 species with diverse life histories. The species visiting T. st ns r nge in body sizes, flor l us ge p tterns, nd nesting h bits. The b l nce of guilds tt r cted, proportion of the n tive bee community (10%) ttr cted, high tempor l t tr ctiveness, nd nnu l blooming suggest th t T. st ns might be n ppropri te t rget resource th t c n highlight trends in the loc l bee community. V ri bilit y in the over ll ttr ctiveness of individu l trees h s been est blished, nd in dividu l cities h ve been shown to h ve different communities of bees. The impli c tions for s mpling, monitoring, nd conserv tion in rel tion to these trends re discussed in further det il. P tterns in resource v ri bility The popul tion of Tecom st ns ex mined showed v ri bility in bee ttr ction. Th is to some extent c n be expl ined by the gre ter l ndsc pe content within which subs mples were found, suggesting th t the surrounding l ndsc pe imp cts the st ructure of urb n bee communities. At finer resolution, individu l resources wi thin the three cities were found to v ry signific ntly in both the number of bee s nd the number of bee groups ttr cted. Between-resource v ri bility in bee vi sit tion h s lso been documented in other loc l species in the f mily Bignoni c e e, n mely, T bebui rose (Fr nkie et l. 1982) nd T. ochr ce . The s me is s een in some members of the f mily F b ce e within this region, including D lberg i retus , Gliricid sepium, nd Andir inermis (Fr nkie et l. 1997; Fr nkie et l. 2005 ). Suggested c uses include: the qu ntity nd qu lity of bloom; v ri t ion in both the volume nd sug r content of nect r (Kev n 2001); inconsistencies in se son l phenology, especi lly dv nced or del yed blooming (Fr nkie et l. 2005 ); v ri bility in second ry compounds found in nect r (Roubik et l. 1995); competition for pollin tors in the surrounding flor l community; nd the ch r c teristics of the surrounding urb n m trix (C ne et l. 2006). The for ging str tegies of bees re optimized, b l ncing resource cquisition wi th competition nd energy consumption. This optimiz tion results in the developm ent of p tterns nd techniques to f cilit te efficient resource cquisition, inc luding: tempor l flight p tterns, feeding speci liz tion, tr p-lining beh vior, nd resource p rtitioning by l ndsc pe loc tion nd vertic l str tum. While vert ic l str tific tion h s been suggested (Roubik et l. 1982; Roubik 1993; Gumbert nd Kunze 1999), in this p rticul r system there were no signific nt p tterns w ith respect to resource height or pl nt form. Inste d, the number of flowers th t resource possessed, me sure of resource bund nce, w s signific nt drive r of bee visit tion. Resources with more flowers ttr cted more bees, irrespecti

L ndsc pe context nd loc l bee communities Ch r cteristics of the surrounding urb n m trix m y be structuring bee visit tio n t different sp ti l sc les. B g ces nd Liberi h ve the most simil r p ttern s of bee visit tion, nd both re situ ted within simil r region l l nd use: c t tle gr zing in modified gr ssl nds punctu ted by n tive dry forest. Ne rby re t he biologic l reserve systems of Lom s B rbud l nd P lo Verde. C s is different , sitting in the center of gricultur l intensific tion. On finer sc le, B g c es nd Liberi re interspersed with more stre ms nd rip ri n re s th n C s. A more refined species-specific n lysis of the bee communities would reve l if t here re more subst nti l trends in t xon presence nd bsence in urb n re s lo c ted within different l ndsc pes. An initi l study of urb n l ndsc pes cross C liforni conducted by Fr nkie et l. (2009b) indic ted th t the species ttr ct ed to set of t rgeted orn ment l pl nts differed between individu l cities, ev en between those loc ted within simil r ecoregions. The community of bees t loc lity is determined by history, context, nd curre nt m n gement pr ctices. Historic l presence/ bsence records re not lw ys v i l ble for site, nd this is the c se for the cities of B g ces, C s, nd Liber i . The bee species th t occur in urb n l ndsc pes c n be mix of loc l n tives migr ting in from the wildl nd, gener lists th t re successful in novel l ndsc pes, or introduced exotics ( common species group in urb n re s due to the in fluence of ctive nd p ssive hum n introductions). P tterns in richness re lik ely driven by the spor dic occurrence of r re species. Attempts to more dequ te ly c t logue species richness will require more extensive s mpling over fourto five-ye r period, due to both n tur l v ri bility in pollin tor popul tions ( see E rdley et l. 2006 nd the 2007 NRC report St tus of Pollin tors in North A meric ) nd l rge-sc le cycles of se son l bee v ri bility, which re common to the tropics (Fr nkie et l. 2005 nd 2009 ). Finer-sc le investig tions of urb n systems Urb n l ndsc pes h ve been shown to be ccept ble nd ppropri te h bit ts for s ome species of bees. The diversity nd import nce of bee communities found withi n cities present opportunities for conserv tion nd m n gement. The m n gement o f ecologic l systems, however, requires n intim te knowledge of ecosystem-level function nd species-specific biology. Further comp risons between different ur b n l ndsc pes nd within the heterogeneous urb n m trix re lso necess ry. Stu dies th t ex mine more th n one l nd-use type re limited in the liter ture. Onl y four studies of urb n bees h ve considered multiple cities (Dreist dt et l. 1 990; Porrini et l. 2003; Fr nkie et l. 2005 ; Fr nkie et l. 2009 ), nd only one w s met - n lysis th t ttempted to integr te inform tion cross multiple l ndsc pes (C ne 2005). Urb n l ndsc pes do sh re m ny simil r ecologic l ch r cteristics in terms of l nd-use p tterns nd disturb nce regimes, but they re lso situ ted within diffe rent ecologic l regions. It is uncle r if there re l rger congruent trends in b ee ecology between different urb n l ndsc pes, nd if these trends re glob lly consistent. The v ri bility in the p tchwork of l nd-use types th t m kes up the urb n l ndsc pe must lso be ex mined in more det il to determine if sm ll-sc l

ve of their height nd pl nt form. This trend w s driven, in p rticul r, by the visit tion of Afric n honey bees (AHB) nd H lictids. The visit tion r tes of Ce ntris nd stingless bees (STB) were not influenced by resource ch r cteristics nd structure, which is consistent with wildl nd studies of for ging in these spe cies groups (Roubik et l. 1982). Contr ry to the results presented here, Roubik et l. (1982) did not est blish ny rel tionship between flor l bund nce nd b ee for ging. The size of the resource p tch w s, however, shown to correl te wit h bee visit tion, with l rger p tches receiving more visits, in studies of crop fields in Engl nd (Cresswell nd Osborne 2004).

e, loc l f ctors influence the success of bees in modified l ndsc pes. Sm ller b odied species such s bees respond to sc le-depend nt f ctors, nd n ppropri t e underst nding of bee ecology within ny l ndsc pe requires loc l-sc le investi g tions. The next step in urb n bee for ging ecology will be to outline microh b it t elements nd sm ll-sc le v ri bility within the urb n m trix nd to determi ne their imp ct on bee occurrence nd community structure. Acknowledgments Funding w s provided by the M rg ret C. W lker Fund in p rtnership with the Essi g Museum of Entomology t the University of C liforni -Berkeley. I m very gr te ful to L urence P cker nd J son Gibbs of York University, Toronto, C n d , nd Ric rdo Ay l of UNAM, for their ssist nce with specimen identific tion. I woul d like to th nk L ur Fine nd Me gh n J strebski, who provided field ssist nce nd good comp ny throughout the dry se son. Addition lly, the hospit lity of th e S ndov l-Argon f mily m de the rese rch experience wonderful.