Patterns of plant invasion along an environmental stress gradient

Introduction

Invasive species are recognized as a major threat to biological diversity worldwide, but our understanding of the dynamics of invasion remains limited (Gaston 2000;Sala et al. 2000). It is particularly unclear what makes a community invasible (Rejmánek 1989;Williamson 1996;Myers & Bazely 2003). Elton (1958) hypothesized that increasing native diversity should confer increasing resistance to invasion, but support for this has been equivocal (Cornell & Lawton 1992;Stohlgren et al. 1999;Kennedy et al. 2002;Fargione et al. 2003). Disturbance is often considered an important precursor to invasion (D'Antonio 1993; Lozon & MacIsaac 1997;Davis et al. 2000) but many undisturbed native communities host exotic flora. Environmental stressors that limit the invader, but not the locally adapted native species, may be important (Harrison 1999), although physically stressful habitats are also not free from the impact of invasive species. Generalities have thus emerged on invasibility but the empirical evidence has yet to definitively support any view in particular.

A limiting factor for untangling the causes of invasibility has been the focus on the most heavily invaded areas and the most successful invaders. However, understanding where invasions do not occur is as important as where they do, as not all communities are invaded equally or affected by the same species. Only by expanding the examination of invasion beyond highly affected areas can we determine if the principles of invasion are contingent on particular conditions. This necessitates broadening our investigations both spatially and structurally, including the examination of community-level patterns of invasion along broad environmental gradients (e.g. Kolb et al. 2002).

We take such an approach in a heavily invaded oak savanna ecosystem in western Canada, examining how patterns of plant invasion change in relation to landscapelevel environmental heterogeneity. This ecosystem is organized along a physical stress gradient of soil depth, elevation, and canopy openness, with shallower soils of higher elevation being more moisture-limited and subject to greater solar exposure due to reduced canopy cover. Although decreasing soil depth is known to reduce above-ground biomass production for both native and exotic species in this system (MacDougall & Turkington 2005), it is unknown whether the composition and abundance of invasives changes with soil depth, whether such factors also impact native plant diversity, and how soil depth interacts with other environmental variables (e.g. elevation, canopy) to shape community structure generally.

We test two hypotheses on the environment-invasion relationship, asking if areas with harsher physical conditions are more or less susceptible to invasion. Areas of higher stress may promote an increase in the richness and abundance of invasives by serving as a refuge from the intense above-ground competition that tends to characterize mesic habitats (e.g. Tilman 1988;Keddy 2001;Grime 2001). Alternatively, such environments may limit invasion by requiring highly localized adaptations that only native species possess (Harrison 1999).

Methods

Study system

The Quercus garryana ecosystem occurs from northern California to southern British Columbia (Fig. 1A), mostly in the lee of the Coastal Mountain range of western North America (Dunn & Ewing 1997;Maret & Wilson 2000;Fuchs 2001). In British Columbia, the climate is sub-Mediterranean, and soils are moderately infertile and of post-glacial origin (Roemer 1972). Soil depths range from more than 1 m to less than a few cm (Erickson 2002). Soils tend to become shallower with increasing elevation, although depth to bedrock can be widely variable in lower elevation areas. Ground flora diversity is high compared to most other plant communities in the region, supporting 454 vascular plant taxa, of which 144 are naturalized exotic tree, shrub, and herbaceous species. Since the early 19th century this ecosystem has become highly fragmented and is represented today by remnant savanna of varying size surrounded by settlements or farms. There are no remnant areas completely free of exotic species. Because many native plant taxa are now uncommon or rare (>10% of total native flora), the oak savanna of southwestern British Columbia is a hot spot for rare floral diversity and one of Canada's most endangered terrestrial ecosystems (Anon. 2002).

Survey procedures

The study was done in seven oak savanna remnants in the Cowichan Valley of southeastern Vancouver Island (Fig. 1B). These areas vary in size, soil depth, elevation, and aspect (Table 1). All contain relatively diverse assemblages of native species. Highly altered remnants were avoided. Remnant size ranged from 0.3 to 18 ha. Five of the sites occur at higher elevations (100-300 m), where shallower soils are generally more prevalent. Most high-elevation sites have aspects ranging from south to west, which maximizes exposure to sun and wind. Less exposed areas (e.g., north-facing hillsides) are dominated by Pseudotsuga menziesii forest that does not support oak savanna species. Six of the sites occur in the eastern end of the Cowichan Valley (Fig. 1B). The seventh site is 15 km to the west on a rock outcrop formation along the Cowichan River. It has become isolated by long-term infilling of Pseudotsuga menziesii in surrounding areas with deeper soils.

Ground flora abundance was determined at each site using 1-m 2 plots located in savanna openings, oak understorey, and transition areas dominated by mixtures of oak and Pseudotsuga menziesii. The number of plots per site was proportional to the size of the remnant (range: 20-38 plots/site, n = 184). Plots were selectively placed to represent opening, understorey, and transition areas; randomly placed plots were not used because they consistently missed rarer microhabitats and, thus, over-sampled highly abundant and wide-ranging species. In each plot, we visually estimated percent cover of all species using a 1-m 2 grid divided into 20 cm × 20 cm cells (Armesto & Pickett 1985). For each species, cover was summed one cell at a time until the entire grid was covered. Six environmental variables were measured for each plot. Aspect and elevation were determined with a GPS unit, accurate to 15 m and calibrated at sea level before elevation readings were taken. Canopy openness was measured at 150 cm from ground level with a hemispherical lens (Solar Pathfinder), with readings ranging from 0 (full canopy cover) to 1 (no canopy). Slope angle was determined with a clinometer. Soil depth to bedrock was the average of four locations per plot, determined using a steel rod driven into the soil. Maximum vegetation height (ground stratum) was measured in each plot as a surrogate for above-ground competitive intensity, recognizing that small differences in stature can have substantial impacts on the abundance and reproduction of subordinate species (Givnish 1982;Gaudet & Keddy 1988;Grime 2001). Species richness was the number of species per plot, and species evenness was calculated using E var (Smith & Wilson 1996).

Data analysis

To determine the relationship between environmental variation and the distribution of all native and exotic species, we performed a direct ordination with Canonical Correspondence Analysis (CCA) (PC-ORD -McCune & Mefford 1999). CCA was preferred over other ordination methods (PCA, NMS) because our main interest was in determining relationships among the environmental variables and plant community pattern (ter Braak 1986, Palmer 1993. All environmental variables were transformed to improve normality and homogeneity of variance before analysis. Aspect was cosine transformed following Beers et al. (1966); canopy openness was arc-sin-square-root transformed. Elevation, slope, maximum vegetation height, and soil depth were log-transformed. CCA calculates two score sets: weighted averaging (WA) scores and linear combination (LC) scores. We used WA scores because they are less sensitive to variations in community-level data that can obscure species-environment relationships (McCune & Grace 2002). We identified and tested the significance of potential outliers in the multivariate space using PC-ORD's outlier analysis. Eight outliers were detected and removed. To test the null hypothesis of no relationship between the species and environmental data matrices, we conducted a 999-run Monte Carlo randomization procedure.

To test the impact of environmental variation on various compositional and functional indices (e.g. evenness [E var ], %annuals), we performed regression analyses using the environmental variable or variables identified by the CCA as most influential on species distributions. Linear fits were used except when a quadratic better captured the relationship. Regressions were conducted with JMPIN (Anon. 2001). Table 1. Summary of environmental data from the seven sites (A-G) used in this study. All data except area and total richness are plot averages. All data except area and total richness are plot averages (+-1 SE). E var (evenness) calculated from Smith & Wilson (1996); 'canopy openness' ranges from 0 (closed canopy) to 1 (no canopy); 'Height' is the maximum canopy height in each plot, and serves as a surrogate for above-ground competitive intensity (Grime 2001

Results

Composition and species richness

We observed 119 plant species from the seven sites, 46 of which were naturalized exotics. Species richness averaged 53 species per site (Table 1). There was no relationship between site area and species richness per site (site richness derived by pooling all plots in each site; r 2 = 0.25, F 1,6 = 1.6, p = 0.26). The largest site (Site C ; Table 1), which has the deepest soil on average, had the smallest number of species per unit area reflecting the absence of a number of native and exotic taxa that only occur in sites with shallower soils.

The most abundant exotic functional group was annual forbs (38% of all non-native species). Perennial and annual forbs were the most abundant natives (41% and 24% of all native species, respectively). There were no native annual grasses observed, although eight native species from this functional group occur elsewhere in southwestern British Columbia (Fuchs 2001). Overall, the most frequently occurring species were the native perennial forbs Camassia quamash (50% of all plots), Lomatium utriculatum (46%), and Triteleia hyacinthina (45%). The annual forb Galium aparine was the most widely occurring non-native species (56% of all plots) and Cytisus scoparius was the most common exotic woody species (41%); 74 of the 119 observed species were rare, with a plot frequency of < 5%.

Relationship between environmental variation and the distribution of plots and species

The proportion of variance explained by the first three CCA axes was small (8.8%), as would be expected given the large number of species, plots, and environmental variables. However, the ordination was significant (P < 0.001 for each axis -Monte Carlo randomization test) indicating that the CCA provided a reasonable summary of the relationships between the species distributions and the six measured environmental variables (Table 2; Fig. 2A, B). The CCA results indicated that the native and exotic species are similarly arrayed in relationship to the measured environmental variables (Fig. 2B). The distributions of species varied widely in response to soil depth, canopy openness, elevation, and vegetation height but there was no indication that exotics responded differently to these variations as compared with the native species. The first CCA axis was most strongly and positively associated with soil depth. It was also positively correlated with vegetation height, and negatively correlated with elevation, canopy openness, and slope (Table 2). This resulted in plots with deeper soils, less canopy openness, and taller ground vegetation being well separated in ordination space from plots at higher elevation with shallower soils and less canopy cover. The second CCA axis was moderately correlated with canopy openness (Table 2). Aspect did not have a strong correlation on any axis. Because soil depth showed the strongest correlation with the first axis, and because canopy openness, elevation, and vegetation height covary with soil depth in the study area (e.g. open canopies only form on shallow soil, high elevation sites only possess shallow soil, the tallest vegetation occurs on the deepest soils), we used soil depth as the main independent variable for subsequent regression analyses of patterns of functional group distribution and compositional indices.

Relationship between soil depth and plant community parameters

Total species richness per plot (native and exotic) was not associated with soil depth (F 1,183 = 0.17, p = 0.68). By contrast, there was a significant curvilinear relationship (F 2,183 = 4.59, p = 0.013) (Fig. 3A) between native richness and soil depth, with richness declining in the shallowest and deepest soils. This was explained by extremely shallow soil (< 3 cm) plots having mostly bryophytes and invasive annual grasses (Aira spp.), while deeper soil (> 50 cm) plots were dominated by invasive perennial grasses (Poa pratensis, Dactylis glomerata, Alopecurus pratensis). On deeper soils, total cover by the invasive grasses averaged > 65% per plot. The few native species that occurred on deeper soils were functionally identical to the dominant invasive perennials, and included the tall-statured (>100 cm height) native perennial grasses Elymus glaucus, Bromus carinatus, Bromus sitchensis, Melica subulata, and Festuca rubra. Total cover by these native grasses averaged > 10% per plot on deeper soils. There was no relationship between exotic species richness and soil depth (F 1,183 = 2.55, p = 0.112) (Fig. 3B).

Declines in native species richness with soil depth were correlated with significant increases in maximum vegetation height (Fig. 3C) and the occurrence of exotic perennial grasses (Fig. 4). There was a significant decline in the percentage of annual species per plot (native and exotic) with soil depth (F 1,183 = 13.52, p = 0.0003) (Fig. 3D). The increasing dominance by the perennial grasses with increasing soil depth was indicated by the significant decline of evenness as the soils became deeper (F 1,183 = 9.39, p = 0.0025) (Fig. 3E). No relationship was detected between native species richness and the number of exotic species (F 1,183 = 1.18, p = 0.279) Fig. 4. Distribution of exotic herbaceous species along the soil depth and canopy openness gradients. These gradients capture the range of habitats sampled in this study, from open areas with shallow soil to closed-canopy sites with deep soil. Italicized lettering identifies perennial grasses widely introduced for pasture enhancement: AO = Anthoxanthum odoratum; PP = Poa pratensis; DG = Dactylis glomerata; HL = Holcus lanatus; AP = Alopecurus pratensis. Errors bars = ± 1 SE. (Fig. 3F). We also detected no relationship between soil depth and the cover of leguminous species (F 1,183 = 0.004, p = 0.95), lessening the possibility of an effect due to N-fixation and soil fertility.

The inverse correlation between plant height and native species richness could conceivably be an artifact of plant size -plots on deeper soils have larger plants so there should be fewer individuals per plot and, by implication, fewer species (Oksanen 1996). To test this, we randomly pooled the 1-m 2 plot data into larger samples of 2 m 2 , 4 m 2 , and 8 m 2 for every 10 cm of soil depth (e.g. 0-10 cm, 11-20 cm) (Rapson et al. 1997). If plant size explains diversity in the 1-m 2 plots, then diversity should increase on deeper soils as plot area increases. This did not occur. Native species richness declined significantly with soil depth in all cases (2 m 2 : r 2 = 0.88, F 2,6 = 11.20, p = 0.035; 4 m 2 : r 2 = 0.81, F 2,6 = 9.7, p = 0.047; 8 m 2 : r 2 = 0.86, F 2,6 = 12.7, p = 0.018).

There was an almost complete change in the composition of exotic species from open sites with shallow soil to closed canopy sites with deeper soil (Fig. 4). Although exotic perennial grasses dominate many oak savanna remnants in British Columbia, their distributions were skewed significantly towards deeper soils (Fig. 4) (F 1,183 = 46.4, p < 0.0001; Tukey's HSD test). Shallow soils were mostly invaded by annual grasses (Aira spp., Vulpia bromoides, Bromus spp.), annual forbs (Galium aparine, Trifolium dubium, and Myosotis discolor), and a small number of perennial forbs mostly from the Asteraceae family (e.g. Crepis capillaris, Hypochaeris radicata) (Fig. 4).

Discussion

General trends

The environmental stress gradient of this savanna ecosystem strongly influenced patterns of plant invasion but not in ways that were predicted. There was no change in the number of invasive species with environmental variation or with the richness of native species. However, the identity of the invaders changed with increasing soil depth, as did their apparent impact on native diversity. There was an almost complete turnover in exotic species composition from shallow to deep soils. Annual grasses, annual forbs, and a few perennial forbs mostly dominated shallow soils, while perennial grasses dominated deeper soils. This pattern of turnover (i.e. ß-diversity) generally matched changes to native species composition along the gradient (for instance fewer native annuals, more native tall-statured perennial grasses in deeper soils). There also appeared to be substantial differences in the impacts of the invaders on native plant diversity. As soil depth increased, the number and abundance of native species declined in association with increasing dominance by a small number of invasive perennial grasses (Poa, Dactylis, Alopecurus). Thus, the hypothesis that invasibility is determined by environmental stress was not supported. Although the soil depth gradient affected the identity of the invaders and their likely impacts on native diversity, all areas were susceptible to establishment by exotic taxa.

The most likely mechanism underlying these patterns is increasing above-ground competition with increasing soil depth. Plots with soils > 50 cm deep are characterized by tall and dense grass swards, with light levels < 2% full light at ground level, and by significantly higher above-ground biomass (MacDougall & Turkington 2005). Experimental disturbance treatments and seed additions have demonstrated that many exotic and native species that are currently limited to shallower soils are fully capable of establishing and regenerating in the absence of these grasses (MacDougall 2005). It appears, therefore, that the segregation of species between deep and shallow soils is driven in part by intense competition in more mesic areas, similar to observations from other systems (Terborgh 1973;Gurevitch 1986;Baskin & Baskin 1988;Keddy 2001). These relatively mesic areas are not more invasible per se but more susceptible to invasion by species with strong competitive impacts on community structure, as has been observed elsewhere (D'Antonio 1993;Milberg et al. 1999;Larson et al. 2001;Meekins & McCarthy 2001;Kolb et al. 2002). Although more xeric habitats (shallow soils) were equally invaded, it does not appear that exotic species in these areas are limiting the distribution or abundance of co-occurring native species.

These results beg the question of why the impact of exotic species appears to diminish in shallower soils. Does the competitive effect of the exotics decrease with increasing stress, possibly due to a home-site advantage by the locally adapted natives? Or does competition become less relevant due to the highly limiting physical conditions (extreme summer moisture deficits and solar exposure) that characterize these areas? The shallow soils do appear to restrict or prevent the growth of the exotic perennial grasses, likely because these species flower in mid-summer when moisture availability can be close to nil (except Anthoxanthum, which flowers March-May). Many native species are adapted phenologically and physiologically to these conditions, finishing their annual reproductive cycle by late spring (May) or by possessing tolerance mechanisms to low moisture and high temperatures (e.g. Opuntia, Sedum).

However, local adaptations are unlikely to be the only cause of reduced exotic impact. Numerous invaders in west-coast oak savanna ecosystems of North America are known to proliferate in conditions similar to the shallow soils of British Columbia (e.g. Bromus tectorum, Centaurea spp.). This suggests that recruitment barriers (geographic or fortuitous), climatic thresholds (too far north), or the variety or subspecies that happened to be introduced could partially explain presentday invasion patterns. Invasiveness and the impacts of invasives on native species, therefore, may be unpredictable based solely on factors such as native diversity or local environmental conditions. The traits and competitive strategies possessed by the arriving pool of invaders, and the historical circumstances underlying their arrival, may also be relevant for determining where invasion occurs and its level of ecological impact.

Trait similarities between exotic and native species

Our results suggest that the functional characteristics of native species in any given habitat (e.g. tall perennial grasses in deep soils) tend to predict the traits of the invaders of those same areas. In deep soils, exotic perennial grasses currently dominate areas where similarly robust native grasses (e.g. Elymus glaucus, Bromus carinatus) most commonly occur. Native and exotic annual forbs tend to occupy shallow soils where exotic perennial grasses are absent. Levels of invasion by the exotic shrub Cytisus scoparius are matched by expansion of the native shrub Symphoricarpos albus in some locations. This result is contrary to recent experimental work demonstrating that invasion success is likely to favour species that are functionally distinct from those in the recipient community (Fargione et al. 2003;Godfree et al. 2004).

In our study area, there is a strong negative association between the occurrence of exotic perennial grasses and native forbs, while the association between the occurrence of exotic and native perennial grasses is strongly positive. Although the native grasses are less abundant than the invaders, they possess a range of life history strategies (e.g., larger seed mass, rapid seedling growth) that allows them to co-exist (MacDougall & Turkington 2004). The experimental removal of the dominant exotic grasses confirms the importance of functional dissimilarity for invasion in this ecosystemthe species that respond greatest to the removals are those most functionally distinct (annuals, forbs, shrubs) (MacDougall & Turkington 2005). The causes of these patterns are unclear but we suggest they further highlight the significance of environmental filters for determining invasibility. It appears that invasion is a two-tier process in this system, where environmental stressors first determine community membership (exotic and native), and competitive interactions then determine relative abundance among species that pass through these filters. What remains unclear is why exotics emerge as the most abundant species within any given habitat, to which numerous hypotheses have been proposed (Williamson 1996;Callaway & Aschehoug 2000;Mack et al. 2000;Myers & Bazely 2003).

Conservation implications

Our results indicate that the effectiveness and urgency of invasive species management are likely to vary along environmental gradients, highlighting the need for ecosystem-based approaches to invasion control. In this oak savanna ecosystem, controlling invasives on deeper soils is a priority due to their apparent competitive impacts on native flora. Of the 59 plant species listed as provincially or nationally endangered in this ecosystem, 27 are associated with deeper soils where the effects of invasive species are undoubtedly exacerbated by habitat loss and long-term fire suppression .

Complicating the control of deep-soil exotics, however, is the large and functionally diverse pool of nonnative species that occurs in the region. Many exotics currently found on shallow soils proliferate following perennial grass removal in areas with deeper soils (MacDougall & Turkington 2005). Exotics such as Bromus sterilis and Trifolium dubium, for example, dominate the deep-soil seed bank and increase threefold or more following disturbance. The shrub Cytisus scoparius similarly proliferates with the removal of the invasive grasses. Although some native taxa also respond positively to these changes (Cardamine oligospora), exotic species typically dominate due to higher propagule pressure, longer seed viability, or faster growth rates in post-disturbance contexts (e.g., MacDougall & Turkington 2004). It seems, therefore, that there is no simple strategy for controlling exotics in this system. For every management action, there appears to be one or more species capable of aggressively responding to its effects. Managers must thus target the most problematic species and also account for the likely after-effects of their removal, including the possible emergence of different but similarly pernicious invaders that are presently confined to areas of higher environmental stress.