Assembly of nonnative floras along elevational gradients explained by directional ecological filtering Jake M. Alexandera,1, Christoph Kueffera,b, Curtis C. Daehlerb, Peter J. Edwardsa, An��bal Pauchardc, Tim Seipela, and MIREN Consortiuma,b,d,e,f,g,h,i,2 aInstitute of Integrative Biology, Swiss Federal Institute of Technology (ETH Zurich), CH-8092 Z��rich, Switzerland bDepartment of Botany, University of Hawaii, Honolulu, HI 96822 cLaboratorio de Invasiones Biol��gicas, Universidad de Concepci��n and Institute of Ecology and Biodiversity, Casilla 160-C, Concepci��n, Chile dDepartamento de Ecolog��a, Universidad de La Laguna, La Laguna, 38206 Tenerife, Spain eDepartamento de Bot��nica, Universidad de Concepci��n and Institute of Ecology and Biodiversity, Casilla 160-C, Concepci��n, Chile fDepartment of Environment, Climate Change and Water, Queanbeyan, NSW 2620, Australia gUSDA Forest Service, Pacific Northwest Research Station, La Grande, OR 97850 hDepartment of Land Resources and Environmental Sciences, Montana State University, Bozeman, MT 59717-3120 and iNational Herbarium of Victoria, South Yarra, VIC 3141, Australia Edited by Harold A. Mooney, Stanford University, Stanford, CA, and approved November 10, 2010 (received for review September 3, 2010) Nonnative species richness typically declines along environmental gradients such as elevation. It is usually assumed that this is because few invaders possess the necessary adaptations to succeed under extreme environmental conditions. Here, we show that nonnative plants reaching high elevations around the world are not highly specialized stress tolerators but species with broad climatic toler- ances capable of growing across a wide elevational range. These results contrast with patterns for native species, and they can be explained by the unidirectional expansion of nonnative species from anthropogenic sources at low elevations and the progressive dropping outofspecieswithnarrowelevationalamplitudes���a pro- cess that we call directional ecological filtering. Independent data confirm that climatic generalists have succeeded in colonizing the more extreme environments at higher elevations. These results suggest that invasion resistance is not conferred by extreme con- ditions at a particular site but determined by pathways of introduc- tion of nonnative species. In the future, increased direct intro- duction of nonnative species with specialized ecophysiological adaptations to mountain environments could increase the risk of invasion. As well as providing a general explanation for gradients of nonnative species richness and the importance of traits such as phenotypic plasticity for many invasive species, the concept of di- rectional ecological filtering is useful for understanding the initial assembly of some native floras at high elevations and latitudes. altitudinal gradient | dispersal | invasibility | nestedness | Rapoport effect Salong everal factors are known to shape species richness patterns elevational gradients, notably energetic constraints on primary productivity and species���area relationships (1, 2). How- ever, these factors are often highly correlated, making it difficult to assign causality, especially because species richness patterns are the result of both contemporary and historical ecological and evolutionary forces. High-elevation floras are typically composed of species with narrow climatic ranges and specialized ecophysi- ological adaptations to low temperatures, such as low stature, slow growth rates, and freezing resistance (3). Because richness gra- dients emerge from the overlap of individual species ranges, some authors have generated null models for richness patterns by as- suming that species ranges are placed at random within a bounded elevational domain (4, 5). This usually produces a mid-domain effect, with richness peaking at mid-elevations where the overlap of species ranges is greatest. Indeed, such mid-elevation peaks do occur, and at least some of them can be explained by the overlap at ecotones of species adapted to different parts of the gradient (6). Although there is a long tradition of studies on elevational richness patterns of native species, little is known about similar phenomena in nonnative species. Nearly 1,000 nonnative plant species have been recorded from mountains throughout the world (7), with species richness usually peaking either at low elevations or toward the middle of the elevational range (8). The decline in nonnative plant richness at higher elevations might reflect the same factors thought to determine richness patterns in native species. A key difference, however, is that although rich- ness patterns of native species have developed over thousands to millions of years, those of nonnative species have assembled from an ecologically diverse pool of species dispersed through human agency over, at most, a few hundred years. For this rea- son, evolutionary factors such as differential rates of speciation (9, 10) may be less relevant for explaining patterns of nonnative species richness than dispersal processes or preadaptation of species to novel abiotic and biotic conditions (11). One reason why it is important to understand the assembly of nonnative species along elevational gradients is that mountain areas, many of them rich in endemic species, remain largely uninvaded (8). The usual explanation is that special adaptations are required to invade extreme environments (8, 12���15), making them inherently resistant to invasion (16, 17). Few studies, how- ever, have explicitly quantified invasion patterns along elevational gradients, although such studies could help explain the apparently low invasibility of extreme environments and provide a basis for their future protection. In addition, understanding the assembly of nonnative species along such gradients could provide insights into the processes determining richness patterns in native floras (18). To reach general conclusions about elevational trends in non- native plant richness and the factors that determine them, we recorded species richness in transects placed at regular elevational intervals along roads in eight mountain regions including five continents and two oceanic islands. All regions were characterized by a steep climatic gradient spanning, on average, a 10 ��C differ- ence in mean annual temperature from bottom to top and a land use gradient from heavily modified lowland to more natural highland habitat. However, several factors differed greatly among regions, including the elevational range, available area, and den- sity of the road network. Our study addressed three core ques- tions. (i) How consistent are elevational richness patterns of nonnative plants around the world? (ii) How are elevational gradients of nonnative plant richness assembled? (iii) Can a Author contributions: J.M.A., C.K., C.C.D., P.J.E., A.P., T.S., and M.C. designed research J.M.A., C.K., C.C.D., P.J.E., A.P., T.S., and M.C. performed research J.M.A. and T.S. analyzed data J.M.A., C.K., C.C.D., P.J.E., A.P., and T.S. wrote the paper and C.K. coordinated the research of the MIREN Consortium. The authors declare no conflict of interest. This article is a PNAS Direct Submission. Data deposition: Georeferenced species occurrence data have been deposited with the Global Biodiversity Information Facility, www.gbif.org (collection code ���MIREN_survey01���). See Commentary on page 439. 1 To whom correspondence should be addressed. E-mail: jake.alexander@env.ethz.ch. 2 The MIREN Consortium members include Jos�� Ar��valo, Lohengrin Cavieres, Hansjoerg Dietz, Gabi Jakobs, Keith McDougall, Bridgett Naylor, R��diger Otto, Catherine G. Parks, Lisa Rew, and Neville Walsh. This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1013136108/-/DCSupplemental. 656���661 | PNAS | January 11, 2011 | vol. 108 | no. 2 www.pnas.org/cgi/doi/10.1073/pnas.1013136108

common mechanism explain elevational gradients in nonnative plant richness in the various regions? Results Two important patterns emerged from our data. First, nonnative species richness always declined from above the lowest one-third to the upper end of the elevation gradients (P 0.0001) (Fig. 1 and Table S1). This pattern was independent of the elevational range and other differences among regions. Richness patterns in the lowest one-third of the gradient were less consistent, with fewer species in the lowest elevation plots in some regions. This het- erogeneity could be explained by a combination of climatic effects ���for example, very dry conditions at the lowest elevations on Tenerife (19)���and patterns of human disturbance. Thus, non- native plant richness was always highest in that part of the eleva- tional range where human activity, using road and population density as proxies (Fig. S1), was greatest. In contrast, native plant richness showed no consistent trend with elevation in the three study regions for which data were available (Table S1 and Fig. S2). Second, the decrease in nonnative species richness with in- creasing elevation was because of a progressive loss of species, and therefore, the species found at high elevations were those with the widest elevational ranges that also occurred at low elevations (Fig. 2). In all regions, the elevational range of species recorded at high elevations was significantly greater (P 0.05) (Fig. 3) than would be expected if ranges were random in re- lation to elevation (4, 21). Additionally, the number of species restricted to the upper one-half of the gradient was significantly smaller than would be expected with random range placement and in five cases, was fewer than three (Fig. 3). Furthermore, a nestedness analysis confirmed that, in five of eight regions, the nonnative species composition of sites was significantly nested in relation to elevation, indicating that the species found at high elevation were a subset of those found at low elevation (Fig. 2). These patterns again contrasted with those of native species, whose elevational ranges were not consistently larger than expected at high elevation (Fig. 4). Furthermore, the proportion of native species found only at high elevations was substantially greater than for nonnative species, and in two regions, the pro- portion was either slightly greater than or not different from that expected with a random placement of species ranges. Discussion In contrast to native species, which tend to be most numerous in the center of the elevational range (although a range of patterns including monotonic decreases have been reported) (1, 22, 23), numbers of nonnative plants consistently peaked at lower ele- vations. Significantly, the richness patterns that we observed were independent of the large differences in the elevational gradients and climates among regions. For example, in both central and southern Chile, species richness declined from ���15���25 species at the bottom to fewer than 5 species at the top, although the ele- vational extents of these gradients did not overlap. This strongly suggests that it is the relative difference in abiotic factors along the gradient from low to high elevation and not region-specific factors such as the available area or particular climatic conditions that drives richness patterns in nonnative plants. Furthermore, the consistency of these patterns around the world suggests that they are explained by a common mechanism. The nesting of species elevational ranges suggests that most nonnative plant species first arrive at low elevations, where an- thropogenic propagule pressure is greatest (8, 23���25), and from there, spread upwards, either naturally or through human agency. Because propagule pressure at high elevations is low, the species that reach the highest elevations must be good dispersers. How- ever, because they are able to establish populations across the full Fig. 1. Global decreases in nonnative species richness with elevation. The relationship between nonnative species richness in plots along roadsides and elevation in eight mountain regions (stars) around the world. For model parameters and statistics, see Table S1. Alexander et al. PNAS | January 11, 2011 | vol. 108 | no. 2 | 657 ECOLOGY SEE COMMENTARY