J. HYM. RES. Vol. 14(1), 2005, pp. 69-77

The North American Invasion of the Giant Resin Bee (: )

Ismael A. Hinojosa-Diaz, Olivia YAnez-Ordonez, Guojun Chen, A. Townsend Peterson, and Michael S. Engel

(IAH, MSE) Division of Entomology, Natural History Museum, and Department of Ecology & Evolutionary Biology, 1460 Jayhawk Boulevard, Snow Hall, University of Kansas, Lawrence, Kansas 66045-7523, USA; (OY) Museo de Zoologia "Alfonso L. Herrera," Departamento de Biologia Evolutiva, Facultad de Ciencias, Universidad Nacional Autonoma de Mexico, Apdo. postal 70-399 CP 04510 Mexico D.F., Mexico; (GC, ATP) Division of Ornithology, Natural History Museum, 1345 Jayhawk Boulevard, Dyche Hall, University of Kansas, Lawrence, Kansas 66045-7561, USA

— Abstract. The giant resin bee, sculpturalis Smith (Megachilidae: Megachilinae), is a species originally of Asia recently adventive in North America. This large and conspicuous species was first recorded at a few localities in the mid-Atlantic states of the United States, but is now found from southeastern Canada (Ontario) to Georgia, and as far west as western Pennsylvania and northwestern Alabama. Known occurrences of this species in its native distributional areas were used to generate an ecological niche model for the species, which can be used to anticipate the geographic potential for species in novel landscapes. The niche model was tested on the native range of the species for robustness in predicting independent suites of occurrence points. The niche model was then used to the distribution of M. in North Amer- — predict potential sculpturalis ica our results indicate that this species has the potential eventually to occupy the entire eastern half of the continent, as far west as the Great Plains. The model also predicts that the species would find appropriate conditions along the Pacific Coast, in Mexico, and in the West Indies. Impacts of M. sculpturalis on native Megachile species are entirely unknown.

As the most significant and efficient pol- lifera Linnaeus native to Africa, Europe, linators of flowering plants, bees are crit- the Middle East, and northwestern Asia ical for many aspects of the diversity and (Ruttner 1988), now globally distributed stability of both natural and agricultural as human colonists have transported bee ecosystems; in addition, honey bees have colonies. A famous episode in the pres- long been appreciated for their products ence of honey bees in the Americas was introduction in 1957 of (e.g., honey and wax; Michener 2000). the experimental These beneficial features make it difficult the African A. mellifera scutellata Lepeletier to think of bees as threats when intro- de Saint Fargeau ("Africanized" honey duced into areas outside their native rang- bees) into Brazil, and the later establish- es, despite the widely known negative ef- ment of feral populations throughout fect of exotic species in general (Goulson South and Central America, reaching the 2003, Lawton and Brown 1986, William- southern United States (Kerr 1957, 1967, son 1999, NAS 2002, Perrings ct al. 2002). Michener 1975, Taylor 1977, Sheppard and Several bee species have been intro- Smith 2000). Goulson (2003) mentioned to im- duced into novel regions by man, either other bee introductions carried out deliberately or not. The most famous ex- prove pollination, among the most signif- ample is the western honey bee Apis niel- icant, species of the genera Bombus, Me- 70 Journal of Hymenoptera Research

Ascher and Brooks gachile, Osmia and Nomia. (2001) polylectic (Mangum 1997). mentioned the presence of 17 adventive Combining the records reported by Batra and bee species in North America, providing (1998), Ascher (2001), and Mangum taxonomic, geographic and biological in- Sumner (2003), in North America, M. formation for Hylaeus (Spatulariella) hyali- sculpturalis has been recorded foraging on natus Smith, and occurrence notes for Au- flowers of at least 16 plants of 12 families, thidium (Anthidium) manicatum (Linnaeus), the most commonly visited being ever- A. (Proanthidium) oblongatwu (Illiger), Hop- lasting pea, Linnaeus Che- litis (Hoplitis) anthocopoides (Schenck), (Leguminosae); Japanese pagoda, Sophora lostoma (Gyrodromclla) rapunculi (Lepeletier japouica Linnaeus (Leguminosae); privet, 1 C. de Saint Fargeau) , (Fovcosmin) campan- Ligustrum lucidum W. T. Aiton (Oleaceae); and ularum (Kirby), (our subject herein) and golden-rain tree, Koelreuteria panicu- Megachile () sculpturalis lata Laxmann (), the first na- Smith. tive to Europe and the remainder to Asia. The resin bee, M. is a giant sculpturalis, Female M. sculpturalis leave a trace of their robust bee distributed in eastern widely foraging activity on flowers of everlasting Asia Taiwan, and (China, Japan, Korea). pea and Japanese pagoda by puncturing The is differentiated from species easily the standard petal (Mangum and Sumner native North American its Megachile by 2003). (14-19 mm in males, 22-27 mm elongate In the last decade, M. sculpturalis has ap- in females) black parallel-sided body, peared in eastern North America, with head, and dark mesosoma with fulvous populations established and spreading setae and Brooks 1997) 1). (Mangum (Fig. from their initial areas of appearance It nests preferentially in shady places, a (probably near Baltimore, accidentally in- minimum of 0.5 m above the ground, in a troduced, via cargo from Japan or China; variety of cavities, e.g., dry, hollow hori- Batra 1998, Mangum and Brooks 1997). zontal stems (bamboo in its native range), Megachile sculpturalis was first collected in and empty burrows made by other hy- North America in 1994 on the campus of menopterans (Iwata 1933, Okada 1995), in- North Carolina State University, and by cluding abandoned wood burrows of car- 1996 was widespread in North Carolina penter bees (Piel 1933). This latter behav- (Mangum and Brooks 1997), also reaching ior has already been documented in North Delaware (Mangum and Sumner 2003). It American populations (Mangum and has since over much of eastern Brooks 1997). Brood cells are made of res- spread North America, with records as far west in from conifers (Iwata 1933) and maple as Athens, Limestone Co., Alabama, as far gum (Piel 1933), from which the name "gi- south as Auburn, Lee Co., Alabama (Kon- ant resin bee" derives (Batra 1998). In Ja- do ct al. 2000), and as far north as Onon- pan, its period of adult activity is from late New York and June through September (Iwata 1933), co- daga Co., (Ascher 2001), Ontario, Canada and Sumner inciding with the blooming of kudzu (Mangum Records also exist from [Pueraria lobata (Willdenow) Ohwi (Legu- 2003). Georgia, South Penn- minosae)], its principal source of pollen Carolina, Virginia, Maryland, West (Batra 1998), although it is known to be sylvania, Virginia, Ohio, Washing- ton D.C., Tennessee, and Connecticut (Mangum and Sumner 2003). Batra (1998) 1 This has often been referred to the old- species by predicted, based on its Asian range, that er, but preoccupied, name of C. fuliginosum (Panzer) M. sculpturalis would come to inhabit the (a junior primary homonym in Apis), which was re- to climates placed by C. nigricorne (Nylander), but this itself is a humid, subtropical temperate synonym of C. rapunculi. of the southeastern and mid-Atlantic Unit- Volume 14, Number 1, 2005 71

habitus. Fig. 1. Megachile sbulpturalis Smith, female from Japan, above dorsal habitus, below lateral

ed States, from eastern Texas and Florida, range ecological characteristics provide north to southern New England. excellent predictivity regarding invaded- Herein we have applied methods of eco- range ecological and geographic potential Panetta Suth- logical niche modeling. Extensive previ- of species (Scott and 1993, ous studies have indicated that native- erst et al. 1999, Skov 2000, Zalba ct al. 2000, 72 Journal of Hymenoptera Research

Peterson et al. 2003). Although this ap- mental Panel on Climate Change, native proach does not provide comprehensive resolution 50 x 50 km: http://www.ipcc. in scale be- predictions of geographic range because ch/). To minimize conflicts of other complicating factors (Peterson et tween topographic and climatic data, we al. 2003), the resulting predictions never- conducted analyses at an intermediate res- X theless offer an excellent summary of spe- olution (10 10 km). — cies' invasive potential. As such, we use Ecological niche modeling. Ecological this technique to predict the potential ex- niches are herein defined as the set of con- tent of M. sculpturalis' invasive range in ditions under which a species is able to North America. maintain populations without immigra- tion Our METHODS (Grinnell 1917, 1924). approach consisted of three steps. (1) Model ecolog- Input data. —Collections with specimens ical niche requirements of the species of M. sculpturalis were studied to obtain based on known occurrences in the native native-range occurrence data suitable for distribution area of the species. (2) Test the retrospective georeferencing. Specimen accuracy of the native-range predictions data were taken from the Snow Entomo- based on spatially structured subsets of logical Collection, Division of Entomolo- the available information. (3) Project the gy, University of Kansas Natural History niche model onto North America to iden- to in- Museum, Lawrence, KS, USA; Kyushu tify areas predicted to be susceptible University, Japan; Institute of Zoology, vasion. Chinese Academy of Sciences, People's The software tool used for niche mod- Republic of China; and the Natural His- eling was the Genetic Algorithm for Rule- tory Museum, London, UK; as well as set Prediction (GARP) (Stockwell and No- data from Huan-li Xu personal collection ble 1992, Stockwell and Peters 1999). (People's Republic of China). Occurrences GARP uses an evolutionary-computing of the species on its invaded range in approach to carry out a flexible and pow- North America were gathered from the erful search for non-random associations Snow Entomological Collection, Universi- between environmental variables and ty of Kansas Natural History Museum, known occurrences of species, as contrast- Lawrence, KS, USA, and from recent pub- ed with the environmental characteristics lished reports (Mangum and Brooks 1997, of the overall study area. Batra 1998, Kondo et al. 2000, Ascher 2001, Specifically, available occurrence points Mangum and Sumner 2003). are resampled with replacement to create To summarize ecological variation a population of 1250 presence points; an across the native and introduced geo- equivalent number of points is resampled graphic distributions of the species, we from the population of grid squares ('pix- used 15 raster grid data sets ('coverages'). els') from which the species has not been These coverages summarized aspects of recorded. These 2500 points are divided topography (elevation, topographic index, equally into training (for creating models) slope, and aspect, from the US Geological and testing (for evaluating model quality) Survey's Hydro-IK data set, native reso- data sets. Models are composed of a set of lution 1 X 1 km: http://edcdaac.usgs. conditional rules developed through an it- gov/gtopo30/hydro/) and climate (an- erative process of rule selection, evalua- nual means of diurnal temperature range; tion, testing, and incorporation or rejec- frost days; precipitation; maximum, mini- tion. First, a method is chosen from a set mum and mean monthly temperatures; of possibilities (e.g. logistic regression, solar radiation; wet days; and vapor pres- bioclimatic rules, etc), and applied to the sure; for 1960-1990 from the Intergovern- training data set. Then, a rule is developed Volume 14, Number 1, 2005 73

by a number of means (mimicking DNA expected under random models. Because evolution: test point mutations, deletions, results depend critically on how oc- to crossing over, etc.) maximize predictive currence points are divided into training accuracy. Rule accuracy is evaluated via and testing data sets (Fielding and Bell the testing data, as a significance param- 1997), we used a 2 X 2 checkerboard ap- eter based on the of percentage points cor- proach (splitting available points into rectly predicted as present or absent by quadrants above and below median lati- the rule. The change in predictive accura- tude and median longitude) that presents cy from one iteration to the next is used a maximum challenge to the model—pre- to evaluate whether a particular rule diction into broad areas from which no oc- should be incorporated into the final rule- currence information was available (Peter- set. The algorithm runs either 1000 itera- son and Shaw 2003). Ecological niche tions or until addition of new rules has no models based on localities in two of the effect on predictive accuracy. The final quadrants ("on-diagonal") were used to rule-set (the ecological niche model) is predict the distribution of the occurrences then projected onto a digital map of native in the other two quadrants ("off-diago- or potentially invaded areas to identify a nal"), and vice versa. Models were vali- potential geographic distribution. Al- dated via chi-square tests that incorporate though these environmental variables can- dimensions of correct prediction of both not represent all possible ecological-niche presences (based on independent test dimensions, they likely represent (or are data) and absences (based on expected fre- correlated with) many influential ones in quencies) (Peterson and Shaw 2003). Ran- delineating the species' potential distri- dom expectations were calculated as the bution. product of the proportional area (within Spatial predictions of presence and ab- 500 km of known occurrences) predicted sence can hold two types of error: omis- present and the number of test presence sion (areas of known presence predicted points. Observed frequencies of correct absent) and commission (areas of known and incorrect predictions of presence were absence predicted present) (Fielding and then compared with expectations using a 2 Bell 1997). Because GARP does not pro- x test (1 df). duce unique solutions, we followed re- KtbU L b cently published best practices approaches to identifying an optimal subset of result- The native-range predictions based on ing replicate models (Anderson et al. the two independent spatial subsets of the 2003). For each analysis, we developed 100 available occurrence data were closely replicate models; of these models, we re- similar to one another (Fig. 2), with the tained the 20 with lowest omission error, exception that the model based on on-di- Finally, from these 20, we retained the 10 agonal quadrants was somewhat more ex- with moderate commission error (i.e., we tensive in the north and the south. Both discarded the 10 models with area pre- predicted areas in the 'other' quadrants dieted present showing greatest devia- (from which occurrence data were with- tions from the overall median area pre- held from the modeling exercise) that co- dieted present across all models). This incided well with the test points in those ; " ! 'best subset' of models was summed to areas (both x * 23.90, both P < 10 "")- produce final predictions of potential dis- Although the two reciprocal predictions tributions. are not identical, their substantial signifi- To validate our model predictions, we cance nonetheless indicates clear predic- for distribution- evaluated their ability to predict indepen- tive ability of our models dent sets of test points compared with that al phenomena related to this species. 74 Journal of Hymenoptera Research

Fig. 2. Predictions of native geographic distribution of Megachile sculpturalis Smith, based on two distinct subsets (here depicted as squares versus circles) of the available data-on-diagonal quadrants predict off- diagonal quadrants (top), and vice versa (bottom). Volume 14, Number 1, 2005 75

Fig. 3. Projection of native-distribution ecological niche model for Megachile sculptumlis Smith to the United of first States, identifying areas putatively suitable for the species (top); observed pattern of advance (year = = detection) of invading populations (bottom): X's 1994-1997, light grav circles 1998-1999, and dark gnu circles = 2000-2001.

Thus, we combined all native-range oc- These areas included the entire eastern as the currences to build a single model for pro- half of the continent, west as far jection to North America. Great Plains. The models also identified distribution Projecting our native-range ecological disjunct areas of potential niche model to North America, broad ar- along the Pacific Coast; in western, cen- eas were identified as suitable for this spe- tral, and southern Mexico; and in the West the distribution cies to maintain populations (Fig. 3). Indies. As such, potential of Hymenoptera Research 76 Journal

in America become established and widespread on of M. sculpturalis North ap- these islands. pears to be quite extensive. The growing number of collection re- DISCUSSION cords of the bee in North America so far confirms the predictions presented here. The invaded distribution in American potential Indeed, plotting known North North America by our ecologi- the predicted occurrences by year (Fig. 3) shows cal niche model for M. sculpturalis consid- of occurrence broadening spatial pattern exceeds the earlier of Ba- erably prediction of the species through time. Continued tra Both that the will (1998). predictions agree surveying of this species over time will come to the humid our species occupy provide much more concrete tests of eastern United States; our model, howev- in states such as predictions, particularly also shows a broader to the and er, potential Kentucky, Illinois, Indiana, Missouri, west and north of the area outlined Ba- local- by Mississippi as well as more western In other instead of ex- tra (1998). words, ities in Tennessee. Such surveys will be west to eastern Texas, the of tending only important for documenting the spread will reach northwestern Tex- for M. species likely M. sculpturalis. Surveying sculptur- western Kansas. instead as and Similarly, alis is not difficult, given its impressive in southern New the of ending England, size; even indirect records can be accu- reach north to Nova species will likely mulated via the marks it leaves on flowers Canada Scotia, (Fig. 3). while foraging (Mangum and Sumner of our model onto all of Projection 2003). Possible impacts of M. sculpturalis America identified additional native North po- on native Megachile species, other areas of invaded distribution for un- tential bees, or as a pollinator, are entirely the eastern the bee outside present range known. Although it has been seen using the discussed above. These areas are along abandoned nests of native carpenter bees Pacific Coast of the United States, portions (Mangum and Sumner 2003), its foraging the of Idaho, and parts of Mexico and activities have so far been recorded prin- Indies These For the moment West (Fig. 3). predictions cipally on exotic plants. —of are—it should be remembered poten- there is little reason to a priori suspect any tial distributions only; as such, the species barriers to the bee's continued westward if would likely establish populations only invasion of the North American continent. with opportunities presented dispersal ACKNOWLEDGMENTS that would place them in or close to those suit- areas. Given the potential ecological We would like to thank Wan-ru Yu (Chinese Acad- traffic of Osamu ability identified and high import emy of Sciences, People's Republic China), Huan-li Xu, from Asia along the western seaboard Tadauchi (Kyushu University, Japan), and Christine (The Natural History Museum, there would, however, appear to be a high Taylor all of whom occurrence data North London, UK), provided of an independent for ad- probability in Asia. We are grateful to Wyatt Mangum introduction of this into was American species ditional information on M. sculpturalis. Support DBI-0096905 S. Ashe northern California, and eastern Oregon partially provided by NSF (to J. and M. S. This is contribution 3401 of the Di- and Washington. The West Indies also Engel). vision of Entomology, Natural History Museum and seems to be a credible region into which Research Center, University of Kansas. related Biodiversity M. sculpturalis might invade. A species, M. (Callomegachile) rufipennis (Fa- LITERATURE CITED bricius) from Africa, is already adventive Anderson, R. P., D. Lew, and A. T. Peterson. 2003. Mitchell in the West Indies (e.g., 1980, distri- Evaluating predictive models of species' Genaro 1997), reinforcing the notion that, butions: Criteria for selecting optimal models. 162: 211-232. if introduced, M. sculpturalis could easily Ecological Modeling Volume 14, Number 1, 2005 77

S. 2001. a Euro- Ascher, J. Hylaeus hyalinatus Smith, Invasions of Nonindigenous Plants and Plant Pests. pean bee new to North America, with notes on National Academy of Sciences, Washington, other adventive bees (Hymenoptera: Apoidea). D.C.; xv + 194 pp. the Proceedings of Entomological Society of Washing- Okada, I. 1995. Behavioral characteristics of the giant ton 103: 184-190. leaf-cutter bee, Megachile sculpturalis. Honeybee Batra, S. W. T. 1998. Biology of the giant resin bee, Science 16: 111-112.' Megachile sculpturalis Smith, a conspicuous new Perrings, C, M. Williamson, E. B. Barbier, D. Delfino, in TJie Naturalist S. immigrant Maryland. Maryland Dalmazzone, J. Shogren, P. Simmons, and A. 42: 1-3. Watkinson. 2002. Biological invasion risks and the A. H. and F. Bell. 1997. Fielding, J. A review of meth- public good: An economic perspective. Conserva- ods for the assessment of errors in prediction tion Ecology [e-publication: journal title recently conservation presence /absence models. Environ- changed to Ecology and Society] 6: http://www. mental Conservation 24: 38-^49. ecologyandsociety.org/vol6/issl/artl/index.html A. 1997. to the Genaro, J. Key genus Megachile, Chal- Peterson, A. T., M. Papes, and D. A. Kluza. 2003. Pre- icodoma group (Hymenoptera: Megachilidae) in dicting the potential invasive distributions of Cuba. Revista de Biologia Tropical 44/45: 193-198. four alien plant species in North America. Weed Goulson, D. 2003. Effects of introduced bees on native Science 51: 863-868. Annual Review and A. T. ecosystems. of Ecology System- Peterson, and J. J. Shaw. 2003. Lutzomyia vectors atic* 34: 1-26. for cutaneous leishmaniasis in southern Brazil: 1917. Field tests of theories Grinnell, J. concerning Ecological niche models, predicted geographic distributional control. American Naturalist 51: distributions, and climate change effects. Inter- 115-128. national Journal of Parasitology 33: 919-931. 1924. and evolution. 5: Grinnell, J. Geography Ecology Piel, O. 1933. Recherches biologiques sur les Hymen- 225-229. opteres vulnerants du Bas Yang-tse. Apidae (les K. 1933. the Iwata, Studies on nesting habits and par- Megachiles). Deuxieme partie (1). Notes d'Ento- asites of Megachile sculpturalis Smith (Hymenop- mologie Chinoise 11: 1-11. tera, Megachilidae). Mushi 6: 4-26. Ruttner, F. 1988. Biogeography and of Hon- Kerr, W. E. 1957. Introducao de abelhas africanas no eybees. Springer-Verlag; Berlin, Germany; Brasil. Brasil Apicola 3: 211-213. xii+284 pp. W. E. 1967. The of the introduction F. Kerr, history of Scott, J. K. and D. Panetta. 1993. Predicting the Africanized bees to Brazil. South African Bee jour- Australian weed status of southern African nal 39: 3-5. plants. Journal of Biogeography 20: 87-93. Kondo, T., M. L. Williams, and R. Minckley. 2000. Sheppard, W. S. and D. R. Smith. 2000. Identification Giant resin bees! Exotic bee species makes its of African-derived bees in the Americas: A sur- way from east coast to Alabama. Alabama Agri- vey of methods. Annals of the Entomological Soci- cultural Experimental Statioti Highlights of Agricul- ety of America 93: 159-176. tural Research 27(3): 5-6. Skov, F. 2000. Potential plant distribution mapping H. K. Lawton, J. and C. Brown. 1986. The population based on climatic similarity. Taxon 49: 503-515. and community ecology of invading . Phil- Stockwell, D. R. B. and 1. R. Noble. 1992. Induction osophical Transactions of the Royal Society of London of sets of rules from distribution data: A B 314: 607-617. robust and informative method of analysis. Mangum, W. A. and R. W. Brooks. 1997. First records Mathematics and Computers in Simulation 33: 385- of Megachile (Callomegachile) sculpturalis Smith 390. (Hymenoptera: Megachilidae) in the continental Stockwell, D. R. B. and D. P. Peters. 1999. The GARP United States. Journal of the Kansas Entomological modeling system: Problems and solutions to au- Society 70: 140-142. tomated spatial prediction. International Journal of Mangum, W. A. and S. Sumner. 2003. A survey of the Geographic Information Systems 13: 143-158. North American range of Megachile (Callomegach- Sutherst, R. W., G. F. Maywald, T. Yonow, and P. M. ile) sculpturalis, an adventive species in North Stevens. 1999. CLIMEX User Guide—Predicting the America, journal of the Kansas Entomological Soci- Effects of Climate on Plants and . CSIRO ety 76: 658-662. Publishing, Collingwood, Victoria, Australia; Michener, C. D. 1975. The Brazilian bee problem. An- iv + 88 pp. L nual Review of Entomology 20: 399^16. Taylor, O. R. 1 >77. The past and possible future Michener, C. D. 2000. The' Bees of the World. Johns spread of Africanized honeybees in the Ameri- Hopkins University Press, Baltimore, MD; cas. Bee World 58: 19-30. xiv + [l]+913 pp. Williamson, M. 1999. Invasions. Ecography 22: 5-12.

T. B. the I. Mitchell, 1980. A Generic Revision of Mega- Zalba, S. M., M. Sonaglioni, and C. J. Belenguer. chiline Bees of the Western Hemisphere. North Car- 2000. Using a habitat model to assess the risk of olina State University, Raleigh, NC; 95 pp. invasion by an exotic plant. Biological Conserva- NAS [National Academy of Sciences]. 2002. Predicting tion 93: 203-208.