Ecological Applications, 19(2), 2009, pp. 359���375 �� 2009 by the Ecological Society of America Multivariate forecasts of potential distributions of invasive plant species INES IBANEZ,1,3 JOHN A. SILANDER, JR.,1 ADAM M. WILSON,1 NANCY LAFLEUR,1 NOBUYUKI TANAKA,2 AND IKUTARO TSUYAMA2 1 Department of Ecology and Evolutionary Biology, University of Connecticut, Storrs, Connecticut 06269-3043 USA 2 Forestry and Forest Products Research Institute, Tsukuba, Ibaraki 305-8687, Japan Abstract. The fact that plant invasions are an ongoing process makes generalizations of invasive spread extraordinarily challenging. This is particularly true given the idiosyncratic nature of invasions, in which both historical and local conditions affect establishment success and hinder our ability to generate guidelines for early detection and eradication of invasive species. To overcome these limitations we have implemented a comprehensive approach that examines plant invasions at three spatial scales: regional, landscape, and local levels. At each scale, in combination with the others, we have evaluated the role of key environmental variables such as climate, landscape structure, habitat type, and canopy closure in the spread of three commonly found invasive woody plant species in New England, Berberis thunbergii, Celastrus orbiculatus, and Euonymus alatus. We developed a spatially explicit hierarchical Bayesian model that allowed us to take into account the ongoing nature of the spread of invasive species and to incorporate presence/absence data from the species��� native ranges as well as from the invaded regions. Comparisons between predictions from climate-only models with those from the multiscale forecasts emphasize the importance of including landscape structure in our models of invasive species��� potential distributions. In addition, predictions generated using only native range data performed substantially worse than those that incorporated data from the target range. This points out important limitations in extrapolating distributional ranges from one region to another. Key words: Berberis thunbergii Celastrus orbiculatus distributional range Euonymus alatus hierarchical Bayes invasive species land cover land use multiscale multivariate New England, USA. INTRODUCTION Invasive species constitute one of the major threats to native organisms and the natural ecosystems they inhabit (Wilcove et al. 1998). As a result, there is great incentive to be able to predict the establishment and spread of invasive species across the landscape. Howev- er, the task of generating accurate predictions is fraught with difficulties. Most predictions are based on knowl- edge of the climate in which a species is already present and then extrapolating to other locations. However, often data are inadequate (i.e., presence-only data or false absences [random or niche-based generation of absence data]) and in most cases, because invasive species are far from equilibrium with their new environment, information based on their current distri- bution may not represent their full potential for spread. As a consequence, predictive models are difficult to calibrate and evaluate for reliability. The appropriate spatial scale and set of explanatory variables seem to vary from species to species and from one invaded region to another. Indeed there has been little attempt to examine invasion processes across multiple scales (e.g., Meyerson and Mooney 2007). To address these uncer- tainties we have developed potential distribution fore- casts of invasive plant species using a comprehensive approach that takes into consideration a broad suite of explanatory variables and scales that may influence plant invasions. Currently, most predictions of future spread of plant species use a ������climate envelope������ approach (e.g., Lasch et al. 2002, Matsui et al. 2004, Hijmans and Graham 2006) in which projected future distributions are based on the current climate in the species��� native range. But species distributions are defined by the complex interactions of biotic and abiotic factors (e.g., Soberon and Peterson 2005, Soberon 2007). Although at times climatic regimes may be the only source of information, forecasts from climate envelopes are limited they ignore the complexity of historical and environmental factors that also contribute to a species��� distributional range (e.g., Broennimann et al. 2007, Soberon 2007, Moles et al. 2008). For example, landscape structure has played an important role in the spread and colonization patterns of invasive plants (e.g., Chabrerie et al. 2007, Melbourne et al. 2007, Theoharides and Dukes 2007). Beyond Manuscript received 20 December 2007 revised 12 June 2008 accepted 30 June 2008. Corresponding Editor: R. A. Hufbauer. 3 Present address: School of Natural Resources and Environment, University of Michigan, Ann Arbor, Michigan 48109 USA. E-mail: iibanez@umich.edu 359

introducing new species, humans have aided plant invasions by creating corridors for their dispersal and by promoting disturbances that may be key to their establishment (see reviews by Lin et al. 2006, Vila et al. 2007). Current distributions of invasive species frequent- ly reflect historical land use patterns (Vila et al. 2003, Domenech et al. 2005, DeGasperis and Motzkin 2007 E. S. Mosher, J. A. Silander, Jr., and A. Latimer, unpublished manuscript), emphasizing the role of human activities in the spread of invasive species. Therefore, incorporating the configuration of the landscape should be a priority in models forecasting future invasions. Finally, explanatory variables are not necessarily simply additive (Pearson et al. 2004, Meynard and Quinn 2007) climate, land use, and other habitat factors will likely interact to affect the establishment of alien species. Dependable forecasts of future spread of invasive species will then be only achieved through a comprehensive approach that incorporates at each relevant scale the major factors shaping the invasion process. In the case of the New England states of the northeastern United States, one-third of the vascular plant flora is nonindigenous, of which 3���5% are invasive (Mehrhoff 2000). These alien species constitute a major threat to the preservation of the natural regional vegetation (Farnsworth 2004) and significantly affect the local economy (Barton et al. 2004, Wang et al. 2006). Despite growing concern and attempts at control, management of known aggressive invaders has not kept pace with the number of new incursions into the northeastern United States, nor indeed elsewhere in the world (Mack et al. 2000, Mehrhoff et al. 2003, Westbrooks 2004, Stokes et al. 2006, Herron et al. 2007). Most states do not have funds to control the majority of the known populations, let alone detect new ones. Thus, given the limited financial and human resources, optimizing early detection and control efforts is crucial to prevent further damage from invasive species. For that, managers need dependable predictions to inform them where these species are most likely to establish, grow, and become focal points for further spread. Our focus here is then to forecast where, in the New England region, three invasive plant species could establish and increase in abundance as successful invaders. We do not intend to predict the propagule pressure and movement patterns that lead to successful invasions, a very complicated task on its own. Instead we want to ask, if propagules are available, could a particular invasive species establish and successfully colonize the area? And then, what are the characteristics of areas that have the potential of being invaded? For that we have developed predictable, spatially extensive outcomes that can be adapted or modified for use in invasion systems elsewhere. We focus on three spatial levels, incorporating the factors affecting invasive species establishment and spread at each scale. First, at a regional level, our predictions will reflect the broad tolerance limits of each species to the regional climate. Second, at the landscape level, taking into account the structure and composition of the landscape, our analyses will inform us about those land use attributes that promote the growth and spread of the invaders. Third, at the local level, local site characteristics, such as local habitat/community type and canopy openness, will inform the in situ establishment conditions that favor colonization of invasive species. We made use of three extensive presence/absence data sets, one from the invaded region and two from part of the native range, a region climatically similar to New England. These constituted one of the most complete presence/absence records used to estimate potential distributional ranges of plant species. We aim to answer the following questions: (1) What are the climatic limits for the spread of our focal species across the region? (2) Given the current climate, which are the most vulnerable areas in the region? (3) What types of land uses facilitate invasions? (4) Which landscape structures are more resistant to invasions? (5) At a local scale, which are the habitats favored by particular invasive species? By addressing all these issues simultaneously we expect to produce reliable forecasts of potential distributions, thus providing direct guidelines for the early detection and management of three invasive plant species. METHODS Study region The region of New England extends .160000 km2 from 4185 0 N to 47829 0 N and from 66854 0 W to 73845 0 W. Plants introduced here encounter a wide variety of climatic conditions and landscapes (Appendix: Figs. A1 and A2). Forests currently cover more than three- quarters of the region the vegetation ranges from oak��� hickory forest in the south to spruce���fir forest towards the north, interior, and higher elevations. In southern New England most forests have regenerated over the past 150 years from pastures and croplands (Foster 1992 E. S. Mosher, J. A. Silander, Jr., and A. Latimer, unpublished manuscript). Forests in the far north are also the result of regeneration after successive rounds of logging activities (Dibble et al. 1999). The introduction of alien species into the region has been common since the first European settlements (ca. 1620). Many of these plants were actively planted for economic and orna- mental uses (Herron et al. 2007). Species analyzed To implement our model of potential distribution of invasive plant species, we selected three woody invasive species common in New England for which we have abundant data. One species from the Berberidaceae family, Berberis thunbergii DC, Japanese barberry, and two species from the Celastraceae family, Celastrus orbiculatus Thunb., oriental bittersweet, and Euonymus alatus (Thunb.) Siebold (synonym E. alata), winged euonymus or burning bush (see Plate 1). These are INES IBANEZ ET AL. 360 Ecological Applications Vol. 19, No. 2

species native to East Asia, being very minor compo- nents in the local communities where they are found (Table 1). All were introduced in the United States in the last half of the 19th century (Mehrhoff et al. 2003) and were reported to be naturalized early in the 20th century. Although their spread is still an ongoing process (Silander and Klepeis 1999, Mehrhoff et al. 2003), we believe the field data we have gathered is sufficient to allow modeling of future expansion across the region. The three species can be found in habitats ranging from full sun to full shade, and they all produce bird- dispersed fruits, indicating that they are habitat gener- alists and can be rapidly dispersed. And, unlike in their native range, the three species show considerable local dominance in the New England plots (Table 1). Berberis thunbergii is a deciduous shrub native to Japan and China, where it typically grows across a wide variety of soil and light conditions in sparse, scattered populations in the forest understory or occasionally larger populations in open sites (Yamanaka 1975, Jarvis and Helin 1993, Mogi et al. 2000). It has spread across the northern half of the eastern United States and has been reported in all the New England states, being very common in the southern and western parts of the region (Fig. 1, Table 1). Although it seems to colonize new areas after disturbances, especially after agricultural abandonment (E. S. Mosher, J. A. Silander, and A. Latimer, unpublished manuscript), it is highly shade tolerant and can form dense thickets under the forest canopy, out-competing the native vegetation (Silander and Klepeis 1999, Mehrhoff et al. 2003). Celastrus orbiculatus is a deciduous woody vine native to East Asia (Japan, Korea, and China), where it occurs most commonly along forest edges, but is occasionally found in the forest interior (Mogi et al. 2000 J. A. Silander, Jr., personal observation). It grows across all of the eastern United States. Its presence in the region is concentrated in the south and the coastal areas, with a few northern incursions (Fig. 1, Table 1). It favors forest edge habitats with full sun exposure, but C. orbiculatus can be also found under dense shade (Leicht-Young et al. 2007). It grows upon any type of support, producing major damage to native plants by girdling and by causing branches to break under its weight (Mehrhoff et al. 2003). Euonymus alatus is a deciduous shrub that can grow up to 2.5 m. It is native to northeastern Asia (Japan, Korea, and central to northeastern China), where it tolerates a wide variety of soil and light conditions and is a fairly common understory shrub (Yamanaka 1975, Mogi et al. 2000). In the United States it is found from northern Florida to Maine and west to Iowa. In New England it is mainly found in the southern part of the region (Fig. 1, Table 1). It can grow in a variety of soil TABLE 1. Occurrences of invasive species across sampled releve plots in their native range in Japan (Phytosociological Releves Database [PRDB] N. Tanaka and I. Tsuyama, unpublished data) and across the New England, USA, plots (Invasive Plant Atlas of New England [IPANE]). Species Plots present (%) Plots with given cover classes (%)�� ,1% 1���5% 6% Japan�� New England�� Japan New England Japan New England Japan New England Berberis thunbergii 2.37 21.1 93.6 5.05 5.2 37.4 1.2 57.5 Celastrus orbiculatus 11.2 21.6 86.9 3.51 9.2 24.8 3.8 71.7 Euonymus alatus 16.9 6.98 81.9 4.76 13.6 38.77 5.5 56.4 �� Percentage of plots with the species present for this cover class. �� Percentage of 7500�� releve field plots sampled with given species present. �� Percentage of 4207 field plots sampled by IPANE volunteers with given species present. FIG. 1. Invasive Plant Atlas of New England (IPANE) presence/absence data for each of the three study species at each of the 4207 locations included in the analysis. Presences are marked by red circles absences are marked by yellow circles. March 2009 361 FORECASTING SPREAD OF INVASIVE SPECIES