Ecological Applications, 17(3), 2007, pp. 663���674 �� 2007 by the Ecological Society of America PREDICTING INVASION RISK USING MEASURES OF INTRODUCTION EFFORT AND ENVIRONMENTAL NICHE MODELS LEIF-MATTHIAS HERBORG,1,2,5 CHRISTOPHER L. JERDE,3 DAVID M. LODGE,2 GREGORY M. RUIZ,4 AND HUGH J. MACISAAC1 1 Great Lakes Institute for Environmental Research, University of Windsor, Windsor, Ontario, N9B 3P4 Canada 2 Department of Biological Sciences, University of Notre Dame, Notre Dame, Indiana 46556 USA 3 Centre for Mathematical Biology, Department of Biological Sciences, University of Alberta, Edmonton, Alberta, T6G 2E9 Canada 4 Smithsonian Environmental Research Center, P.O. Box 28, Edgewater, Maryland 21037 USA Abstract. The Chinese mitten crab (Eriocheir sinensis) is native to east Asia, is established throughout Europe, and is introduced but geographically restricted in North America. We developed and compared two separate environmental niche models using genetic algorithm for rule set prediction (GARP) and mitten crab occurrences in Asia and Europe to predict the species��� potential distribution in North America. Since mitten crabs must reproduce in water with 15�� salinity, we limited the potential North American range to freshwater habitats within the highest documented dispersal distance (1260 km) and a more restricted dispersal limit (354 km) from the sea. Applying the higher dispersal distance, both models predicted the lower Great Lakes, most of the eastern seaboard, the Gulf of Mexico and southern extent of the Mississippi River watershed, and the Pacific northwest as suitable environment for mitten crabs, but environmental match for southern states (below 358 N) was much lower for the European model. Use of the lower range with both models reduced the expected range, especially in the Great Lakes, Mississippi drainage, and inland areas of the Pacific Northwest. To estimate the risk of introduction of mitten crabs, the amount of reported ballast water discharge into major United States ports from regions in Asia and Europe with established mitten crab populations was used as an index of introduction effort. Relative risk of invasion was estimated based on a combination of environmental match and volume of unexchanged ballast water received (July 1999���December 2003) for major ports. The ports of Norfolk and Baltimore were most vulnerable to invasion and establishment, making Chesapeake Bay the most likely location to be invaded by mitten crabs in the United States. The next highest risk was predicted for Portland, Oregon. Interestingly, the port of Los Angeles/Long Beach, which has a large shipping volume, had a low risk of invasion. Ports such as Jacksonville, Florida, had a medium risk owing to small shipping volume but high environmental match. This study illustrates that the combination of environmental niche- and vector-based models can provide managers with more precise estimates of invasion risk than can either of these approaches alone. Key words: ecological niche modeling Eriocheir sinensis genetic algorithm for rule set prediction (GARP) introduced range invasive species native range relative risk risk assessment. INTRODUCTION One of the most challenging problems confronting invasion biologists is the accurate identification of locations that may be successfully colonized by nonin- digenous species. This requires knowledge of the number of individuals introduced into a particular area over time (i.e., propagule pressure), as well as measures of environmental suitability (where a species can survive and reproduce, i.e., its ecological niche). Only in areas where both factors are present does the risk of establishment exist. Here we combined ecological niche modeling and the release of ballast water into ports in the United States to estimate the relative risk of establishment of the invasive Chinese mitten crab. This novel approach allows for spatially focused management efforts based upon port-specific risk levels. Ecologists have sought to link invasion success to many individual factors including diversity of native species, intensity of natural or human disturbance, resource availability, habitat productivity, and the release of invading species from natural enemies (see reviews, Lodge 1993, Levine and D���Antonio 1999, Davis and Pelsor 2001, Kolar and Lodge 2002, Levine et al. 2004, Lockwood et al. 2005, Walker et al. 2005). Recent studies have established that introduction effort (i.e., the number of introduction attempts and propagules per attempt) is positively correlated with invasion success (e.g., see Drake et al. 2005, Lockwood et al. 2005, Manuscript received 10 February 2006 revised 11 August 2006 accepted 25 August 2006. Corresponding Editor: J. C. Callaway. 5 Present address: Pacific Biological Station, Fisheries and Oceans, Canada, 3190 Hammond Bay Road, Nanaimo, British Columbia V9T 6N7 Canada. E-mail: HerborgL@pac.dfo-mpo.gc.ca 663

Ruesink 2005, Suarez et al. 2005, Drake and Lodge 2006). Drake and Lodge (2004) applied the introduction effort concept to marine coastal environments by assessing risk based upon global shipping patterns, which were assumed to be directly proportional to the volume of ballast water discharged and the number of individuals introduced. Wonham et al. (2005) explored theoretically how different scenarios of ballast water exchange would influence introduction effort and risk associated with discharges of ballast water in coastal areas. Many other studies have demonstrated that determi- nants of success may be complex, affected by multiple factors simultaneously (e.g., Lonsdale 1999, Rouget and Richardson 2003, Forsyth et al. 2004, Romanuk and Kolasa 2005). For example, Cleland et al. (2004) noted that diversity of nonindigenous plants was related to both native species diversity and resource availability, while Blumenthal (2005) proposed that release from natural enemies could account for the success of invading plant species in high-resource environments. Stage-based approaches have also been used to predict invasion success (Colautti and MacIsaac 2004). Successful establishment of a nonindigenous species depends upon its successful navigation of a series of transitions, each with independent probabilities of failure (Carlton 1985, Kolar and Lodge 2002). A transport vector must deliver a sufficient number of viable and reproductively capable propagules to an area outside of the species��� historic range. These individuals must be capable of surviving ambient physical and chemical conditions, as well as interspecific interactions with residents of the community. Movement between transitions may be influenced by one or more separate or interacting factors (Colautti et al. 2006). Furthermore, different life history attributes (e.g., growth rate, environmental tolerance) may be important for different transitions. For example, Kolar and Lodge (2002) suggested that for freshwater fish species introduced to the Laurentian Great Lakes, those with high growth rates were favored at the establishment phase but performed poorly at the spread stage. Sequential combining of different approaches to predict invasion success may provide insights beyond the capabilities offered by individual approaches. For example, Peterson (2003) observed that areas in California that should be vulnerable to Asian long- horned beetles (Anoplophora glabripennis), based upon current inbound ship traffic from Asian source ports, are inhospitable to the beetle���s environmental needs and thus unlikely to become invaded. Peterson (2003) used genetic algorithm for rule set prediction (GARP) to determine the environmental needs of the species by matching the suite of environmental conditions associ- ated with presence of the species in its native range with the similarity of these environmental conditions in the novel range. Genetic algorithm for rule set prediction is an environmental niche modeling application that has been used to predict range expansion and limitation for a variety of taxa including fish, reptiles, and caribou (e.g., Raxworthy et al. 2003, Johnson and Gillingham 2005, McNyset 2005). It also has been applied to studies that seek to forecast the vulnerability of sites to establishment by nonindigenous species, based upon the degree of environmental matching between the species��� native and nonnative ranges (Peterson and Vieglais 2001, Gane- shaiah et al. 2003, Arriaga et al. 2004, Drake and Bossenbroek 2004, Iguchi et al. 2004, Roura-Pascual et al. 2004, Underwood et al. 2004). The approach may be particularly useful if combined with measures of intro- duction effort collectively they provide managers with information on areas where species are being introduced as well as the environmental suitability of these habitats. The focus of our study was the Chinese mitten crab (Eriocheir sinensis), a species whose native range spans from 218 to 418 N and includes China, Hong Kong, and North Korea. The species is catadromous: its free- swimming planktonic larvae (duration 43���90 days) develop predominantly in saline water and require salinity greater than 15 psu, although they spend much of the remainder of their life cycle in freshwater (Anger 1991). Adult crabs migrate back to estuaries, where they reproduce and die. This lifestyle constrains the range over which the species may establish viable populations to coastal rivers proximal and connected to brackish or salt water. Although the species is occasionally reported in solely freshwater habitats such as the Great Lakes (Nepszy and Leach 1973), there exists no evidence indicating that the crab can complete its entire life cycle in these habitats. The crab initially invaded northern continental Europe between 1912 and 1940 and southern France between 1950 and 1960 (Herborg et al. 2003). It subsequently spread to Great Britain, the Spanish and Portuguese Atlantic coasts, and to Sweden, Finland, Poland, Estonia, and Lithuania proximal to the Baltic Sea (Silfverberg 1999, Valovirta and Eronen 2000, Ferdinand-Martinez and Carrera 2003). While it is inherently difficult to assess the most likely transport vector for a nonindigenous species, ballast water transport seems to be very important for mitten crabs. The initial introduction of mitten crabs into Europe occurred via ballast water (Herborg et al. 2003). Similarly, the most likely introductory pathways of mitten crabs into the United States are ballast water (Gollasch et al. 2002) or intentional introduction via the live seafood trade (Cohen and Carlton 1997). Chinese mitten crabs have a very patchy distribution in North America, with reports of solitary individuals from Lakes Superior, Erie, and Ontario, the Detroit and St. Lawrence Rivers, the Mississippi River delta, and very recently in Chesapeake Bay (Nepszy and Leach 1973, de Lafontaine 2005 G. Ruiz, unpublished manuscript). The only locality known to support established populations LEIF-MATTHIAS HERBORG ET AL. 664 Ecological Applications Vol. 17, No. 3