Northern Prairie Wildlife Research Center
The presence of a three-way interaction in the best model for number and frequency of alien species in aggregate, which did not occur for any individual species (Table 2), suggests that the interaction is an emergent property of the alien species assemblage, which combines soil and water tolerances and the establishment requirements of each individual species. Mixed-grass prairie in this region is thought, based largely on observations in shortgrass and tallgrass prairies (Daubenmire 1978, Burke et al. 1997), to be limited by nitrogen and water; our explanatory variables likely reflect differences in these resources. Roads and trails are often characterized as conduits for the transport and/or dispersal of alien plants (Tyser and Worley 1992, Knops et al. 1995, Kotanen et al. 1998). Some have suggested that soil moisture is enhanced by runoff from paved roadways (McIntyre and Lavorel 1994, Kotanen et al. 1998). Vegetation types themselves vary in available water and nutrients. Andropogon scoparius is a very efficient competitor for nitrogen (Wedin and Tilman 1993), which may limit the nutrient's availability to aliens and thus compromise their ability to invade vegetation types dominated by this species (Fig. 1). Vegetation types that occur near rivers, tributaries, and drainages and on north-facing slopes are more mesic than those on uplands and south-facing slopes (Hanson and Whitman 1938, Whitman and Wali 1975). Although mesic vegetation types at TRNP tended to have greater richness and/or frequency of alien plants (e.g., Symphoricarpos occidentalis/Prunus virginiana, Artemisia cana, and Populus deltoides/Juniperus scopulorum had relatively more alien species at higher frequency than drier Stipa comata/Bouteloua gracilis, Artemisia tridentata/Atriplex confertiflora, and Chrysothamnus nauseosus/Agropyron smithii), disturbance seemed to favor aliens in only four vegetation types (Agropyron smithii/Stipa viridula [number] and Juniperus scopulorum/Oryzopsis micrantha [frequency] in both units, S. occidentalis/P. virginiana [frequency] and P. deltoides/J. scopulorum [number] in the north unit). Compared with A. smithii/S. comata vegetation, in which disturbance did not influence invasibility, water availability is greater in A. smithii/S. viridula (Hanson and Whitman 1938, Redmann 1975). Added moisture along roadways may have allowed those alien species that were more water-limited to establish, and from there disperse into vegetation types where water was relatively more available. The difference between north and south units may reflect the presence of Euphorbia esula in mesic areas of the south unit (Fig. 3f). This highly invasive weed displaces native vegetation (Belcher and Wilson 1989), and also may outcompete many alien species.
Those species that occurred infrequently (with the exception of Salsola iberica) were unlikely to vary significantly with vegetation type (Table 2). There are three potential reasons for this observation. One explanation is simply a matter of sample size: because there were 11 vegetation types, only uncommon species that were concentrated in a small number of these types, such as S. iberica was, could possibly show a detectable effect. Infrequent clonal species, which are typically undersampled by frequency techniques, were likely missed entirely (e.g., Hyoscyamus niger appeared on only one transect, Sonchus arvensis on only four [Appendix]). Two other explanations are more biologically interesting, but these require further investigation. First, it is likely that stochastic events and disturbance are more important early in an invasion (Mack 1995). As a species becomes more abundant, differences in survival among vegetation types will become evident. This scenario predicts that species we found infrequently will either increase or decrease over time. The second of these explanations is that the species that occurred infrequently are limited by the availability of disturbed sites, and/or are at the mercy of stochastic events to arrive at these sites at appropriate times. This scenario predicts that these species always will be encountered at low frequencies and independent of biotic interactions. Control of such species, which are unlikely to pose a threat to undisturbed native communities, would be a lower priority than control of species whose populations are likely to change unpredictably (Hiebert 1997).
Our results are consistent with a large body of research that has demonstrated greater invasion into riparian zones and mesic sites (Rejmanek 1989, DeFerrari and Naiman 1994, Planty-Tabacchi et al. 1996, Kotanen et al. 1998, Stohlgren et al. 1998). Riparian vegetation types (P. deltoides/J. scopulorum and A. cana) and hardwood draws (S. occidentalis/P. virginiana) were among the most heavily infested at TRNP, both in terms of the number of alien species and their frequency on transects (Fig. 1). The high frequency of alien plants in these vegetation types is noteworthy in that it may suggest greater impact to these native communities than, for example, to A. smithii/S. viridula, which also has large numbers of alien species (Fig. 1a) but at a frequency near the average for all vegetation types (Fig. 1b). Two xeric vegetation types, Chrysothamnus nauseosus/Agropyron smithii, which is dominated by bare ground, and Artemisia tridentata/Atriplex confertiflora, had both low numbers of alien species and low frequencies on infested transects (Fig. 1). As Rejmanek (1989) pointed out, however, using survey data it is not possible to separate resistance of the biotic community to invasion from resistance of the abiotic environment, so reasons for the differences in alien richness and frequency among vegetation types cannot be determined from the current study.
Care must be taken in the interpretation of frequency data with respect to ecosystem effects. Although frequency is not necessarily correlated with cover or biomass, it seems intuitively reasonable that the more often a species is encountered in a defined area, the more likely it is that the species may exert an influence on that plant community. However, the shape of the sampled area and the location of sample units are important in interpreting frequency data (Whysong and Miller 1987, Bonham 1989). Our sample units, transects, were long and arrow and likely undersampled clonal invasives with patchy distributions such as Bromus inermis and Cirsium arvense. We thus may underestimate the effects of such species. Likewise, frequency data typically overestimate species that may be sparsely but evenly distributed (Elzinga 1998). Bromus japonicus and Tragopogon dubius may fit into this category. We have minimized these effects by using the randomly located transect, rather than plots on transects, as our sample unit for number of alien species and for individual alien species: a species was either present or absent in these analyses. Where we did use frequency of occurrence on transects, we averaged over all species, which minimized the effect of any one growth form.
Transects excluded from this analysis (Table 1) support the notion that mesic and riparian areas are more invasible than drier vegetation types at TRNP. Alien species were found on all transects in river bottom and Salix vegetation types. Grassed sand floodplain, which is scoured by ice most years, supports fewer alien plants, perhaps because of the especially harsh conditions. Populus tremuloides/Betula occidentalis, another heavily infested community, occurs at the higher elevations of hardwood draws in the north unit, just above the Symphoricarpos occidentalis/Prunus virginiana vegetation type. Many of the excluded vegetation types (e.g., marsh, S. occidentalis, and Populus tremuloides) occur in small patches, which have been shown to be more invasible in other regions (Kemper et al. 1999, Harrison 1999a). Vegetation complexes, which are a fine-grained mosaic of variously dominant species, are more heavily invaded than might be anticipated based on soil and moisture properties. Scoria is clay that has been baked by burning lignite, for example, and would not be considered mesic, yet is highly invaded (Table 1). Stohlgren et al. (1999) suggested that species-rich grasslands may be more easily invaded, in part, because "transient" species (those that occur infrequently) result in incomplete utilization of nutrients at local scales. Such a mechanism may explain the unexpectedly high levels of alien infestation in vegetation complexes at TRNP.
Several authors have asserted that all habitats or communities are invasible (e.g., Crawley 1987, Usher 1988, Williamson 1996). Our data support this assertion; none of the vegetation types we surveyed were free of alien plants (Fig. 1). Williamson (1996: 56) further stated, ". . . all communities are more or less equally invasible". This seems not to be the case in many systems (e.g., Rejmanek 1989, Harrison 1999b, Stohlgren et al. 1999), including mixed-grass prairie at TRNP. Whether one considers susceptibility to invasion of a single species (e.g., Figs. 2 and 3), or to alien species as a group (Fig. 1), substantial variation existed among vegetation types at TRNP. At a larger scale, Stohlgren et al. (1999) found differences in alien plant invasion among various grassland types, ranging from shortgrass to tallgrass prairie, which they related to differences in native species richness, fertility, and grazing history. Harrison (1999a, b), working in serpentine and nonserpentine meadows in California, likewise found differences in alien invasion that she attributed to soil nutrient characteristics. Rejmanek (1989) found evidence for greater invasion in mesic and early successional communities. Our results are consistent with the idea that both water and nitrogen limitation may influence the number of alien plants in different vegetation types at TRNP. Appropriate experiments would be necessary to determine cause and effect, however.
In general, the interaction between vegetation type and park unit reflected shifts between which unit had the highest proportion of a taxon in a given vegetation type, rather than major differences in vegetation types occupied in the north and south units. For example, Poa pratensis was more commonly encountered in the south unit in P. deltoides/J. scopulorum and in the north unit in Agropyron smithii/Stipa viridula; nonetheless, the species was relatively common in both vegetation types in both park units (Fig. 3g). Exceptions to this pattern were Cirsium arvense and Camelina microcarpa, which occurred at vastly different frequencies between units (Fig. 3d, e), and Euphorbia esula (Fig. 3f), which in the north unit is confined to the river bottom. Bromus japonicus and Melilotus officinalis were always more commonly encountered in the south unit than in the north, when they differed, but relative frequency among vegetation types was similar between units (Fig. 3a, b). Because these two species are an annual and a biennial, respectively, and the extensive M. officinalis seedbanks are known to respond dramatically to favorable germination conditions (Turkington et al. 1978), the difference in frequency between the two park units may reflect differences in temperature and precipitation between the two years in which the units were surveyed, rather than conditions specific to the units.
The frequency with which park unit and disturbance were indistinguishable in their effect on species distributions (Table 2) suggests similar underlying processes. A likely commonality is stochastic events that influence propagule availability and conditions for seedling establishment. Stochasticity may act at multiple scales. Disturbance, as we defined it, reflects stochastic events at a local scale, park unit at a regional scale. Mack (1995) emphasized the importance of environmental stochasticity in the early establishment of immigrant species. At TRNP, if park unit does reflect stochastic events, the effect persisted in several alien species that were abundant and seemingly well past the establishment phase (Table 2, Fig. 2). Models for these species also included vegetation type, either as a main effect or interaction with park unit, and park unit as a main effect usually had an AIC score closer to the best model than did disturbance (Table 2). This suggests the possibility of systematic differences between the two park units that maintain plant populations at different levels. Disturbance, on the other hand, mainly influenced distributions of less common species (Table 2, Fig. 4), although Bromus inermis (Fig. 5) was an exception. For these infrequent species, AIC scores for disturbance and park unit differed by no more than one or two points from each other. The ambiguity in these models may reflect inadequate sampling of uncommon species, although a clear distinction was made between vegetation type and the other two main effects.
Communities of large herbivores may account for some of the observed differences between the north and south units. Although bison (Bison bison) stocking rates are similar in the two units, both elk (Cervus elaphus) and feral horses occur only in the south unit. In the absence of most predators, herds of all three species require periodic thinning to protect park habitat. Grazing by native ungulates and feral horses likely decreases the dominance of some native species; in tallgrass prairie decreased dominance has been related to increased invasibility (Smith and Knapp 1999), although the reverse has been shown in shortgrass steppe (Milchunas et al. 1992).
Because we have restricted our analysis of disturbance to only those attributed to human endeavor, we have not accounted for all disturbances that occur in the park. In fact, disturbance as we have defined it may be qualitatively different from natural disturbances (see Hobbs and Huenneke 1992). For example, there are several large prairie dog (Cynomys ludovicianus) towns in both park units and bison activity results in disruption of native sod at both large and small scales in both units. The overall effect of many natural disturbances at TRNP is to produce changes in continuity of the vegetative canopy and thus increase the number of safe sites for seedling establishment. On the other hand, disturbance associated with roads, campgrounds, and the like, more often takes the form of increased density of alien propagules through seeding for stabilization and landscaping and inadvertent introduction of propagules by humans and livestock. In particular, those areas seeded to dense, sod-forming aliens, such as Bromus inermis, probably have fewer opportunities for seedling establishment than many of the native vegetation types. One exception to these differences between natural and anthropogenic disturbance is the Little Missouri River. Riparian areas have been described as dispersal corridors for alien species (DeFerrari and Naiman 1994). In this case, the natural disturbance is not only a source of propagules, but also provides establishment sites through seasonal flooding and ice scouring. This, along with greater access to moisture near the soil surface, likely provides an environment favorable to establishment of many alien species at TRNP.
The effect of anthropogenic disturbance at TRNP was invariant, in that no alien species was significantly more common on undisturbed transects (Figs. 4 and 5). Of those species associated with disturbed transects, two were grasses widely planted either for roadside stabilization or for forage (Agropyron cristatum and Bromus inermis). Thus, their introduction into TRNP was actively promoted through cultivation and the propagule bank was undoubtedly enhanced through this process. Tyser and Worley (1992) similarly found that disturbance, in the form of road corridors, influenced the number of alien species that invaded adjacent native grasslands. They suggested that not only were roadside plantings responsible for high numbers of aliens adjacent to roads, but that livestock contributed to the further dispersal of alien plants along trails and into the back country. Horseback riding is permitted at TRNP and concessioners' horses have been allowed to roam the south unit, perhaps contributing to the broad distribution of palatable alien grasses throughout the park.
Two other species that were more common on disturbed transects in at least one park unit, Convolvulus arvensis and Chenopodium album, are widely distributed agricultural weeds that are often introduced as contaminants of seed. Because anthropogenic disturbance at TRNP includes areas that have been seeded during construction, these species also may have been planted, albeit inadvertently. Nonetheless, their spread into native vegetation types is associated with proximity to disturbance.
An important outcome of our research is the demonstration that different alien species respond differently to both disturbance and native vegetation type. As Stohlgren et al. (1999) pointed out, sweeping generalizations about alien species are not justified if individual species do not respond in concert to disturbance. We would add that native plant communities vary in their susceptibility to invasion: variation in distributions of the most frequently encountered alien species at TRNP was related to native vegetation type, not disturbance (cf. Tragopogon dubius, Euphorbia esula, Poa pratensis, Bromus japonicus, and Melilotus officinalis, Figs. 2 and 3).