Structural, compositional and trait differences between native- and non-native-dominated grassland patches

Authors


Summary

  1. Non-native species with growth forms that are different from the native flora may alter the physical structure of the area they invade, thereby changing the resources available to resident species. This in turn can select for species with traits suited for the new growing environment.
  2. We used adjacent uninvaded and invaded grassland patches to evaluate whether the shift in dominance from a native perennial bunchgrass, Nassella pulchra, to the early season, non-native annual grass, Bromus diandrus, affects the physical structure, available light, plant community composition and community-weighted trait means.
  3. Our field surveys revealed that the exotic grass B. diandrus alters both the vertical and horizontal structure creating more dense continuous vegetative growth and dead plant biomass than patches dominated by N. pulchra. These differences in physical structure are responsible for a threefold reduction in available light and likely contribute to the lower diversity, especially of native forbs in B. diandrus-dominated patches. Further, flowering time began earlier and seed size and plant height were higher in B. diandrus patches relative to N. pulchra patches.
  4. Our results suggest that species that are better suited (earlier phenology, larger seed size and taller) for low light availability are those that coexist with B. diandrus, and this is consistent with our hypothesis that change in physical structure with B. diandrus invasion is an important driver of community and trait composition.
  5. The traits of species able to coexist with invaders are rarely considered when assessing community change following invasion; however, this may be a powerful approach for predicting community change in environments with high anthropogenic pressures, such as disturbance and nutrient enrichment. It also provides a means for selecting species to introduce when trying to enhance native diversity in an otherwise invaded community.

Introduction

Some non-native plant species have captured the attention of scientist and land managers because they form visually conspicuous, low diversity stands (Levine et al. 2003; Vilà et al. 2011). Often the rise to dominance of a non-native species coincides with the loss or decline of native species including those that were previously dominant. Non-native species with growth forms that are different from the native flora may alter the structural complexity and habitat heterogeneity of the area they invade, thereby changing the resources available to resident species (Crooks 2002; Molinari, D'Antonio & Thomson 2007; Asner et al. 2008; Jager, Kowarik & Tye 2009). Few studies have compared physical structure between native and non-native-dominated communities and even fewer have investigated the ways in which physical structure can affect the availability of resources and subsequent community composition. In this study, we compared differences in physical structure, resource availability, and associated plant diversity and composition between invaded and adjacent uninvaded habitats to gain insight into mechanisms through which a dominant non-native species affects the environment that it invades.

Recently, there has been a surge in ecological interest in the role of species traits in predicting patterns of species composition, invasion and community structure (Diaz et al. 1999; McGill et al. 2006; Westoby & Wright 2006; Kuster et al. 2008; Molina-Montenegro et al. 2012). Most of the focus has been on traits of invaders and why they are successful (Rejmanek & Richardson 1996; Sutherland 2004; Ordonez, Wright & Olff 2010; Van Kleunen, Weber & Fischer 2010; Dostál, Weiser & Koubek 2012). The traits of species able to coexist with invaders are rarely considered when assessing community change following invasion. Yet an understanding of these traits is critical to predicting which species can persist following invasion and which combinations of species may be successful in restoration where invaders cannot be successfully eliminated. Certain traits may make individuals more susceptible to changes in physical structure, disturbance regime or resource availability following invasion, and these are the traits that should be targeted to predict how species composition will change following invasion (Hobbs 1997). For instance, a non-native species that forms a dense canopy and reduces light availability relative to an uninvaded habitat may select for species that are tall, fast growing or have large thin leaves appropriate for harvesting light in a low light environment. In this study, we calculate community-weighted trait means to examine the response of species to differences in physical structure and resource availability associated with the dominance of a non-native species.

We studied grasslands within the California floristic province that contains a rich native flora that has been supplemented with over 300 non-native species (D'Antonio et al. 2007). Although exotic annual grasses currently dominate a large proportion of grasslands in the state, there are a considerable number of locations where native species, usually dominated by perennial bunchgrasses, occur in patches within invaded settings. These matrices of non-native and native-dominated grassland patches provide an ideal situation for the comparison of changes in physical structure and community composition that accompany invasion. The most common native perennial bunchgrass in California coast range grasslands is Nassella pulchra (purple needle grass), and this species forms discrete bunching patterns across the landscapes where it dominates. In contrast, Bromus diandrus (ripgut brome), a Eurasian annual grass that is widespread throughout California, New Zealand, Australia and Chile (Parsons & Moldenke 1975; Tozer et al. 2007; Kleemann & Gill 2009), germinates early in the rainy season and forms dense aggregations that produce a persistent and thick leaf litter layer (thatch). Differences in physical structure of the vegetation between grasslands dominated by N. pulchra and B. diandrus have not been quantified, but appear to be substantial both in terms of the complexity of the vertical and horizontal live plant and litter biomass. Field studies have shown that dense vegetation and accumulated plant litter can affect species composition by altering the availability of resources, such as light, soil moisture and temperature (Knapp & Seastedt 1986; Facelli & Pickett 1991; Jensen & Gutekunst 2003) which in turn have negative impacts on forb and grass germination and survival (Xiong & Nilsson 1999; Jensen & Meyer 2001; Amatangelo, Dukes & Field 2008). We propose that species composition, both in terms of species identity and traits, will be different in grasslands patches dominated by B. diandrus because of their dense growth early in the rainy season and persistent high litter biomass compared with grassland patches dominated by N. pulchra.

In addition to quantifying the physical structure and composition of these two dominant patch types, we assessed four plant traits: plant height, seed size (length and mass), flowering phenology and specific leaf area (SLA), that we hypothesize might change with invasion in these grasslands. Short statured species may be less common in B. diandrus-dominated patches because the dense seedling growth typical of invasive annual grasses has the potential to reduce light at the soil surface, selecting for taller growing species that can keep their photosynthetic area at or above the canopy level of the grasses. Similarly, large seeds have more resource storage than small seeds giving them an advantage in low light settings such as grasslands with a high density of annual exotic grasses and thick persistent leaf litter (Grime & Jeffrey 1965; Amatangelo, Dukes & Field 2008). In contrast, small seeded species often require light to germinate and as a result may not persist in areas shaded by litter and dense annual grasses (Grime et al. 1981; Gross & Werner 1982; Jensen & Gutekunst 2003). Species that are offset in their phenology may compete less intensely with B. diandrus for below-ground resources; however, we suspect light availability at peak growth to be a more important determinant of species abundance in our system (Dyer & Rice 1999). Therefore, we expect early season annuals that overlap with peak B. diandrus growth and reproduction to be more abundant in invaded grassland patches compared with species that are active later in the growing season when light availability has been reduced by the dense canopy of B. diandrus. Finally, we hypothesize SLA to be greater in B. diandrus-dominated patches because large leaf surface area will facilitate light capture. In addition, SLA is positively correlated with growth rate (Reich, Walters & Ellsworth 1997) and faster growing species will be able to grow quickly through leaf litter and dense vegetation and to keep up with the rapidly growing B. diandrus canopy as it develops. We gathered seed size, plant height, flowering phenology and SLA measurements and used species' relative abundance to calculate community-weighted trait means in adjacent N. pulchra – and B. diandrus-dominated patches to investigate how these traits were represented in native and invaded grasslands.

We used paired transects in grasslands containing a matrix of N. pulchra and B. diandrus patches occurring on similar soil type, slope and aspect, and with the same land use history to investigate the following three questions. First, how does the physical structure of the living vegetation and plant litter differ between grassland patches dominated by the native N. pulchra vs. the exotic B. diandrus and how do these structural differences influence light availability? Second, how does the species composition differ between the two dominant grass species (N. pulchra vs. B. diandrus)? Lastly, how do species traits vary in the invaded and un-invaded grassland patches?

Materials and methods

We worked at the University of California, Sedgwick Reserve, 55 km north of Santa Barbara in the foothills of the Santa Ynez Mountains (34˚40′31·61′N, 120˚2′26·42′W). The reserve is 29 km from the coast, between wetter coastal prairies and the drier grasslands of the Central and San Joaquin Valleys. The climate is Mediterranean, with hot dry summers and cool wet winters. The growing season typically begins with autumn rains (October or November) and ceases in May. Rainfall totals during our study were 167 and 403 mm for the 2006–2007 and 2007–2008 rainfall years, respectively, with the annual long-term rainfall averaging 396 mm. In addition to 2007 being a dry year, the timing of the first substantial rainfall event (>5 mm) was later in the season (December) than in the 2008 rain year (October). Hence, the 2007 growing season was much shorter in duration than the 2008 one.

We haphazardly selected ten grassland sites with similar topography (i.e. flat) and soils (See Appendix S1 in Supporting Information), no history of tillage and located away from the influence of the scattered oak trees (Quercus agrifolia and Quercus lobata). Within each site, a set of paired transects was established with one 30-m transect in N. pulchra dominated grassland (>25% N. pulchra cover) and the other in adjacent grasslands dominated by B. diandrus (>25% B. diandrus cover). All paired transects were within 50 m of one another and all sites were at least 200 m from one another. The transects within each pair were chosen to be as close to one another as possible so that dispersal of species between adjacent patches should not have been an important factor affecting composition. All sites were on an uplifted terrace with Positas Series fine sandy loam soils derived from parent material of the Paso Robles formation (Dibblee 1966). Prior to 1995, all sites were grazed by cattle year round for decades in a high-intensity grazing regime. In 1996, the University of California, Santa Barbara established a winter wet rotational grazing program. Five of our sites (including both transects in each pair) were within the grazing experiment and as a result were grazed from 1996 to 2007. In 2008, grazing began after our surveys were completed.

Physical Structure and Species Composition

In 2008, we randomly selected three 0·5-m2 quadrats at six of the ten sites to quantify vertical and horizontal vegetation structure and light availability at the soil surface. As none of the sites had been grazed in 2008, there were no direct effects of grazing on physical structure. Vertical vegetation structure was measured by averaging the number of times vegetation came into contact with a 1-m tall point-intercept vegetation pole that was dropped vertically every 10 cm along the length of the quadrat. Spatial heterogeneity of vegetation (i.e. horizontal vegetation structure) was measured by using the vertical vegetation structure to calculate the coefficient of variation across the length of the sampling quadrat. Thatch depth was measured in four locations in each quadrat by placing a meter stick on the soil surface and measuring the height of dead vegetation lying on the ground. Light availability was measured by placing a point sensor on the soil surface while simultaneously taking an ambient reading above the vegetation using an AccuPAR PAR-80 light meter (Decagon Devices, Pullman, WA, USA). Vegetation and leaf litter were moved aside during placement of the light sensor and returned to their previous position before the reading was taken. The per cent of available photosynthetic active radiation (PAR) at the soil surface was calculated as the% of ambient PAR above the vegetation.

Vegetation surveys were performed during peak plant growth in the springs of 2007 and 2008 by placing ten 0·5-m2 quadrats at 3-m intervals along each of the paired transects. Per cent cover of each species, bare-ground, thatch and small mammal disturbance were estimated using cover classes. In addition to species richness, the Shannon diversity index was calculated. These two indices of diversity followed identical trends between B. diandrus and N. pulchra dominated patches in both survey years, therefore for brevity and ease of interpretation we report on the patterns associated with species richness.

Data were averaged for all 10 0·5-m2 quadrats along each of the 10 N. pulchra and B. diandrus transects, and a paired t-test was used to evaluate the differences in species richness, Shannon diversity and physical structure (bare-ground, thatch, small mammal disturbance and vegetation complexity) in the two patch types. All data were normally distributed with the exception of thatch depth, small mammal disturbance and native grass richness in 2007 and 2008; therefore, a Wilcoxon signed-rank test was used for comparisons between N. pulchra and B. diandrus patches. All statistical analyses were performed using jmp 9.0 (SAS, Cary, NC, USA). To understand the distribution of species across the landscape in each year, we counted whether an individual species occurred in both N. pulchra and B. diandrus-dominated grasslands or whether the species was constrained to only N. pulchra or only B. diandrus-dominated grasslands.

Trait Analysis

We identified four traits, vegetative height, seed size (seed length without adornment or awn) and mass, first month flowering (months were numbered chronologically from the start of the new year), and SLA (leaf area per leaf mass), that we hypothesized to be important for survival and growth in structurally distinct habitats particularly ones differing in light availability. We used the Jepson Manual (Jepson & Hickman 1993) for vegetative height, seed size and flowering dates and supplemented using other field guides (Smith 1952; Munz 1959), data bases [Ecological Flora Data base (http://ucjeps.berkeley.edu/efc/dbase.html)] and field measurements as needed. SLA was solely obtained via field measurements and calculated on three leaves (from separate individuals) by scanning leaves and analysing the images with Image J (Abramoff, Magalhaes & Ram 2004), petioles were excluded from the analysis. Leaf area for grasses was acquired using the leaf blade (i.e. the portion of the leaf above the sheath and ligule).

A community-weighted mean (CWM) was calculated for each trait in each patch type by multiplying the trait value by the relative abundance for each species at the quadrat scale and summing over all species in a given quadrat. We averaged all 10 quadrats along a transect to obtain a transect level mean and used the CWM at the transect scale as our level of replication (N = 10). We performed a Behrens–Fisher t-test to evaluate whether the CWM for each trait was different in B. diandrus vs. N. pulchra dominated patches. In addition to the pairwise comparisons, we performed a univariate regression analysis to test whether traits were influenced by the abundance of B. diandrus and thatch. For each regression, we used the mean CWM for each trait at the transect scale as the response variable and the% cover of B. diandrus vegetation + thatch as the explanatory variable. Transect was the level of replication (N = 20). We performed a sequential Bonferroni analysis for the Behrens–Fisher t-test and univariate regression to correct for multiple comparisons within year. A Box-Cox transformation (λ = 2) was performed on the 2007 CWM for first month flowering and the natural log was used to transform the CWM of SLA in 2008 to meet the assumption of normality. To verify that CWM traits were not driven by differences in soil properties, we reran each of the regression models from 2008 (the year soil variables were collected) and independently included each significant soil characteristic (% rock, cation exchange capacity, base saturation, Ca, Mg and electroconductivity) as a covariate in the model.

It was not possible to acquire trait data for unidentified species; therefore, all unknown species (N = 1 in 2007 and N = 3 in 2008) were excluded from the analysis, their cover did not exceed 2% in any transect. We also excluded species for which trait data were not available. We were able to obtain height and first month flowering values for all species in our surveys; however, we removed nine species from the seed length analysis, six species in 2007 and seven species in 2008 from the seed mass analysis and 11 species in 2007 and 12 species in 2008 from the SLA analysis. The abundance of the excluded species never exceeded 1·5% cover at the transect scale in 2007 and 2008, with the exception of Trifolium depaparatum var. truncatum (missing value for SLA) with a cover of 19% in one N. pulchra dominated transect in 2008. Bromus diandrus and N. pulchra were also removed from the CWM analyses.

Results

Physical Structure

Bromus diandrus-dominated patches were structurally different from those dominated by the native grass, N. pulchra. Vertical vegetation was twice as continuous in B. diandrus compared with N. pulchra patches (Fig. 1a; t = 7·65, N = 6, P = 0·0006), and the coefficient of variation was lower for B. diandrus than N. pulchra grassland patches (Fig. 1b; t = −2·65 N = 6, P = 0·045), indicating that horizontal structure is more homogenous in invaded patches. Bromus diandrus-dominated patches had deeper thatch (1·25 cm deeper, Z = 3·17, N = 10, P = 0·011), 22% more thatch cover (Fig. 1c; t = 3·56, N = 10, P = 0·0061), 12% less bare-ground (Fig. 1c; t = −6·31, N = 10, P = 0·0001) and 6·5% more small mammal disturbance (Fig. 1c; Z = 25·00, N = 10, = 0·0059) than N. pulchra-dominated patches. There was a threefold decrease in available light at the soil surface in B. diandrus compared with N. pulchra patches (Fig. 1d; t = −9·68, N = 6, P = 0·0002).

Figure 1.

Paired t-test comparing (a) Vertical vegetation structure (# of vegetation hits along a point-intercept pole), = 0·0006, (b) Variation in vertical structure (coefficient of variation calculated using 10 vertical vegetation measurements across each plot), = 0·045, (c)% cover of thatch (= 0·0061), small mammal disturbance (= 0·0059, Wilcoxon signed-rank test) and bare-ground (= 0·0001), (d)% of available photosynthetic active radiation at the soil surface, = 0·0002 in Bromus diandrus (black bars) and Nassella pulchra (grey bars) dominated grasslands. Error bars ± 1 SE.

Species Diversity

Shannon's Diversity Index (2007: t = −2·62, N = 10, P = 0·028; 2008: t = −2·28, N = 10, = 0·049) and species richness (2007: t = −3·92, N = 10, P = 0·0098; 2008: t = −6·44, N = 10, < 0·0001) were lower in grasslands dominated by B. diandrus compared with those dominated by N. pulchra. Native forbs were the most affected by B. diandrus dominance with the average quadrat having 0·75 fewer species in 2007 (t = 5·65, N = 10, = 0·0005) and 3·4 fewer species in 2008 compared with quadrats in N. pulchra patches (Fig. 2). At the quadrat scale, richness of native grasses was higher in N. pulchra patches than in B. diandrus patches in 2007 (Z = −27·50, N = 10, = 0·002) and 2008, while richness of exotic grasses and exotic forbs were not significantly different between the two patches in either year (Fig. 2).

Figure 2.

Mean ± 1 SE for species richness per 0·5 m2 for each of four functional groups [native forb (NF), native grass (NG), exotic forb (EF) and exotic grass (EG)] in 2008. NG and NF richness are significantly different between Nassella pulchra and Bromus diandrus-dominated patches in both 2007 (not shown) and 2008 (NG: Z = −27·50, N = 10, < 0·0001, NF: t = 6·52, N = 10, = 0·001).

We summed all the unique species at the landscape scale (across the 10 paired transects) and found that the majority (59% in 2007 and 53% in 2008) of species occurred in both N. pulchra and B. diandrus patches (Fig. 3). In 2007, there were 16 unique species (12 native and four exotic species) in N. pulchra patches and nine unique species (six native and three exotic species) in B. diandrus patches. In 2008, there were 25 unique species (21 native and four exotic species) in N. pulchra patches and seven unique species (three native and four exotic) in B. diandrus patches (Fig. 3).

Figure 3.

The total number of species that occurred in both Nassella pulchra and Bromus diandrus patches, N. pulchra patches only or B. diandrus patches only across all 10 grassland sites (50 m2). The species are pooled according to their origin (native vs. exotic) and survey year (2007 and 2008).

Traits

The CWM for height showed that tall species were more abundant in B. diandrus-dominated patches, while short statured species were more abundant in N. pulchra patches in both survey years (Fig. 4a, 2007: t = −4·00, N = 10, = 0·0008 and 2008: t = −3·29, N = 10, = 0·002). The CWM for seed mass was higher in B. diandrus-dominated patches than in N. pulchra patches (Fig. 4b, 2007: t = −3·64, N = 10 = 0·0021 and 2008: t = −4·09, N = 10 = 0·0008). Only seed mass [and not seed length (2007: t = −2·50, N = 10, = 0·024 and 2008: t = −1·19, N = 10, = 0·25)] is represented in Figs 4 and 5 for brevity. In 2008, species that flowered earlier in the year were more abundant in B. diandrus-dominated patches while those that had a later flowering phenology were more abundant in N. pulchra patches (Fig. 4c; 2007: t = 0·36, N = 10, = 0·71 and 2008: t = 3·54, N = 10, = 0·002). SLA was not significantly different between the patch types in either year.

Figure 4.

Community-weighted means (CWMs) for each trait in Bromus diandrus (black bars) and Nassella pulchra (grey bars) dominated patches in 2008. *< 0·05, **< 0·01, Error bars ± 1 SE.

Figure 5.

Regression analysis of community-weighted mean for each trait (y-axis) and the sum of Bromus diandrus vegetation and thatch cover (%) (x-axis). Symbols represent survey year, ■ = 2007, ○ = 2008. Transects in both patch types were used for the analysis (N = 20). *< 0·05, **< 0·01, NS = not significant.

Regression analyses suggested that when the CWM for traits differed between patch types (above), the% cover of B. diandrus vegetation and thatch could explain the variation in the CWM. For instance, height was significantly greater in B. diandrus-dominated patches than in N. pulchra patches and the variation in the CWM (for height) was predicted by the cover of B. diandrus and thatch (Fig. 5a). Likewise, the% cover of B. diandrus and thatch was related to seed length in 2007 (R2 = 0·29, = 0·03, N = 20) and seed mass in both years (Fig. 5b). Variation in first month flowering, SLA and seed length (in 2008) was not explained by the% cover of B. diandrus vegetation and thatch. In general, the explanatory power of the regression models did not improve when soil characters were added as covariates, however CEC and Mg were significant predictors of CWM (first month flowering) in 2008 (CEC: R2 = 0·61, < 0·001, N = 20 Mg: R2 = 0·62, < 0·001, N = 20).

Discussion

Grassland patches dominated by the exotic annual grass B. diandrus differ in physical structure, diversity and trait values when compared to patches dominated by the native perennial grass N. pulchra. Many studies have reported a loss of species in invaded habitats; however, very few explore potential causes for this decline (Levine et al. 2003; Vilà et al. 2011). Bromus diandrus alters the physical structure in such a way that available light is significantly reduced at the soil surface. Species that are better suited (taller, earlier phenology and larger seed size) for low light availability are able to coexist with B. diandrus, supporting our interpretation that light availability is likely to be an important driver of lower diversity in B. diandrus relative to N. pulchra patches and that coexistence of forbs within this grassland setting may depend on their traits relative to the dominant grass. However, we recognize that competition for below-ground resources may also contribute to the structuring of these grassland communities and experimental manipulations are needed to determine the role of light relative to other potential drivers of compositional change (e.g. nitrogen, water availability, chemical properties of the litter, etc.).

Physical Structure

Exotic plant species that change physical structure have been proposed to alter diversity of the resident community, and this relationship depends on the trait similarities between the non-native and native dominants (i.e. growth form, phenology, annual production). Crooks (2002) suggests that exotic plants that increase physical structure (i.e. architecture of the vegetation and leaf litter) relative to the pre-invasion dominants are likely to have positive effects on the diversity of the resident species, while those that decrease structure will have the opposite effect. Until recently, these patterns have mainly been based on cross trophic interactions between exotic plants and higher trophic levels (i.e. insects, birds and mammals) and their importance is just beginning to be realized for plant assemblages. For instance, Jager et al. (2009) compared the effects of the non-native tree, Cinchona pubescens, on plant community composition in a formerly treeless habitat vs. its effects in a habitat previously dominated by a native shrub. They found that species composition in the naturally treeless zone was more affected by the invasion of the non-native tree than composition in the formerly shrub dominated habitat where species were more likely to be tolerant of shading.

We found differences in the physical structure of invaded and uninvaded grasslands despite the similarity in stature and life-form (both are grasses) of B. diandrus and N. pulchra. Grassland patches invaded by B. diandrus have complex vertical structure (composed of both living vegetation and thatch) that is homogenous across the landscape with a minimal cover of bare-ground. In contrast, grasslands dominated by N. pulchra were characterized by having dense vertical structure in areas where the bunchgrasses were present and bare-ground in the interstices between the bunches, thereby making these grassland more heterogeneous. Few studies have assessed differences in physical structure between invaded and uninvaded landscape patches, yet physical structure can alter critical plant resources (e.g. light and water), making it an integral part of understanding the underlying drivers of diversity loss following invasion. Similar results to ours were found in (sub)montane rain forests in Hawaii where the exotic trees, Fraxinus uhdei and Morella faya, increased forest biomass by 32–51% which decreased light availability and resulted in very few understory species (Asner et al. 2008). While changes in physical structure following invasion are rarely documented, they may have substantial impacts on resident plant species by altering critical resources, like light.

Reduced light availability in invaded patches relative to native-dominated patches is a frequent mechanism invoked to explain differences in diversity between such patch types (Braithwaite, Lonsdale & Estbergs 1989; Kwiatkowska et al. 1997; Reinhart et al. 2006; Jager, Kowarik & Tye 2009; Saito & Okubo 2011). The impacts of reduced light availability is expected to be particularly high in habitats that have not historically been challenged by low light conditions, such as in semi-arid grasslands dominated by low stature bunchgrasses. Our work corroborates that of Dyer & Rice (1999) which suggests that invasion by annual grasses in California grasslands has shifted the primary limiting resource both spatially (from below-ground water availability to above-ground light availability) and temporally (from summer to winter/spring). Many semi-arid native grassland species may lack traits that are well suited for survival under these new conditions.

Species Composition

We found grassland patches that were invaded by B. diandrus to have fewer species than neighbouring uninvaded patches with the most impacted group being native forbs (>80% were annual species). Many studies have documented the decline in richness following invasion (Levine et al. 2003; Vilà et al. 2011), yet few have investigated species additions and preferences (i.e. higher abundance) in invaded patches compared to native-dominated patches. We found unique species (both native and exotic) that only occurred in invaded grassland patches (Fig. 3), as well as a number of species that were significantly more abundant in invaded patches (see Appendix S2). The majority of species that were favoured in B. diandrus patches were exotic species, perhaps supporting the idea of ‘invasional meltdown’ (Simberloff & Von Holle 1999). There were, however, few native species that were more frequently associated with B. diandrus suggesting that there can be facilitative effects of B. diandrus on a select few native residents. Facilitation by exotic species is usually rare and typically only a few species benefit at the expense of many (Rodrigues 2006).

Trait Variation

We selected traits that we hypothesized would vary between uninvaded and invaded grassland patches. Height, seed size and flowering phenology were consistent with our hypotheses, while SLA was not different between the grassland patch types in either year. We found that small statured species are the most affected by the invasion of B. diandrus and this suggests that changes in physical structure may select for species that are tall and capable of growing through thatch or simultaneous with a homogenous layer of living B. diandrus to reach the sunlight. We also observe many of these tall species (e.g. Amsinckia menziesii and Galium sp.) growing under the canopy of nearby savannah oak trees where light intensity and water availability is moderated by the presence of the oak and its leaf litter. The difference in height in the two patch types may be further exacerbated by the differential investment of N. pulchra and B. diandrus to above-ground vs. below-ground structures. Native species are thought to have higher root–shoot ratios than non-natives (Ehrenfeld 2003), and this was confirmed using native and non-native C4 grasses in the tall grass prairie (Wilsey & Polley 2006), as well as B. diandrus and N. pulchra in California grasslands (Holmes & Rice 1996). Nassella pulchra has a well developed root system and invests more energy in root production than exotic annual grasses (Eviner & Firestone 2007); therefore, other grassland species whose strategy consists of investing in above-ground rather than below-ground growth may be challenged by competition for below-ground resources in N. pulchra dominated patches.

Our finding that large seeded species were more abundant in B. diandrus patches is consistent with other studies where large seeded species are better able to establish in dense vegetation and leaf litter while many small seeded species often require light for germination (Grime et al. 1981; Gross & Werner 1982; Burke & Grime 1996; Jensen & Gutekunst 2003; Kostel-Hughes, Young & Wehr 2005). In addition to B. diandrus altering light availability at the soils surface, its thatch can also create a barrier to seed dispersal which might prevent seeds of lightweight, small seeded species from reaching the mineral soil. We have observed B. diandrus seeds that are trapped in the thatch to develop primary roots that extend 2 cm to the top of the mineral soil. Such growth might be possible for a large seeded species with large nutrient stores, but is unlikely for small seeded species.

Species with earlier flowering phenology were more abundant in B. diandrus-dominated patches in 2008 which was a wet year with early season rainfall in comparison with the 2007 rainfall year. Germination triggers in semi-arid grasslands are often associated with temperature and length of seed maturation (after-ripening) prior to the first substantial rain event of the season (Young & Evans 1989). In years where rainfall starts in autumn and continues at regular intervals, non-native annual grasses, like B. diandrus, are some of the first species to germinate followed by other species that have more complex germination cues like temperature and after-ripening (Gulmon 1992). In 2008, we suspect that the early flowering phenology of grassland species was favoured in B. diandrus patches because B. diandrus became active early in the growing season, while other species with more complex germination cues remained dormant leading to delayed growth and the inability to compete with or survive under already established B. diandrus. Our findings indicate that the timing and amount of yearly rainfall determines whether phenology plays an important role in determining species composition in semi-arid grasslands.

Contrary to our expectation that high SLA would be favoured in invaded patches with low light availability and the need for fast growth, we did not find a relationship between SLA and species abundance in either patch type. Large thin leaves (high SLA) may complicate the ability for a plant to navigate through the dense vegetation and thatch found in B. diandrus patches, thereby negating any positive effect a species with large thin leaves might gain from intercepting more light or growing quickly. In addition, SLA is negatively correlated with seed mass (Maranon & Grubb 1993; Grotkopp, Rejmanek & Rost 2002); therefore, the effects of B. diandrus on SLA may be suppressed because of the positive relationship we found between seed mass and B. diandrus dominance (Fig. 5). To better understand this relationship, we included the CWM of seed mass as a covariate in the regression of the CWM of SLA on B. diandrus and thatch cover and found that the model improved, but was still not significant. SLA has also been shown to track soil properties with high SLA species being favoured in more productive habitats (Cunningham, Summerhayes & Westoby 1999; Poorter & Bongers 2006); however, our work as well as others have found relatively small differences in N availability between soils under native vs. non-native grasses during the growing season (Parker 2006; Corbin & D'Antonio 2011).

Assessing how trait values in a community respond to invasion is a novel approach to measuring community level impact following invasion. Whether these results are generalizable for other systems that have been invaded by a species that alters light availability is yet to be determined and will likely depend on the characteristics of the exotic species (i.e. growth form of the exotic could increase, decrease or create highly variable light availability) as well as the trait diversity of the resident community.

Drivers of Community Change

The majority of studies evaluating diversity loss following invasion use neighbouring native and invaded patches to measure differences in species composition (Vilà et al. 2011). Spatial comparisons, such as these, assume that the lower richness in invaded patches is a direct effect of the invader, yet there may be other underlying factors, such as those that promoted the invasion in the first place, contributing to the disparity in richness between the invaded and native patches (MacDougall & Turkington 2005). We cannot be certain whether B. diandrus was the cause or a consequence of a historic disturbance that led to lower richness in grasslands where it dominates; however, we do know that tillage and grazing, two frequently cited causes of species turnover in grasslands (Barbour, Keeler-Wolf & Schoenherr 2007), were not responsible for the patterns found here because our study sites were never tilled and all the sites were grazed intensively in the past.

Grasslands are also subject to disturbance by small mammals [e.g. pocket gopher (Thomomys bottae), ground squirrel (Spermophilus beecheyi)] which more frequently turn over the soil in B. diandrus patches. If small mammal disturbance was a driver of B. diandrus invasion and species loss, we might expect for these small scale disturbances to further promote B. diandrus and have a negative effect on other species. Yet there is evidence that the opposite is true. Small seeded and low statured species often respond positively to disturbance and bare soil; therefore, we would expect small mammal disturbances in invaded patches to benefit species that are unable to tolerate reduced light levels and physical barriers created by B. diandrus vegetation and thatch (Fehmi & Bartolome 2002).

We measured soil properties (see Appendix S1) that can provide insight into the structuring forces that currently shape species composition in these patches. In the grasslands of California, soil properties frequently define where native and non-native patches occur (Huenneke et al. 1990; Harrison 1999). In our study, we did not find differences in some macronutrients, like N, P, K or water; however, there were differences in the concentration of cations (Mg and Ca) between the patches. These differences could have contributed to both the invasion success and lower richness in B. diandrus-dominated patches, yet, we cannot rule out whether they are in fact the effect of invasion by B. diandrus. We found a twofold decrease in Mg in soils dominated by B. diandrus, but B. diandrus vegetation itself held significantly more Mg above-ground compared with N. pulchra. Thus, the lower level of soil Mg may be an effect of invasion rather than a cause. Similarly, Blank (2010) has found high levels of Mg accumulation in a related invasive Bromus (Bromus tectorum) particularly in contrast to two native perennial grasses in an arid steppe habitat in Nevada, USA. In contrast to Mg, Ca was higher in soils dominated by B. diandrus, but this difference does not appear to be related to differential uptake or differences in the amount of standing biomass in B. diandrus vs. N. pulchra patches suggesting that slight differences in Ca existed prior to invasion. Exotic species are frequently charged with altering the soils in the regions they invade (Ehrenfeld 2003) thus making it challenging to determine whether soil variability between invaded and native patches existed prior to invasion or whether they are an effect of the invasion itself. In the case of our study, differences were slight and may reflect both pre-existing differences and invasion driven differences.

Conclusions

The use of traits to evaluate how communities change following invasion is a new approach to measuring non-native species impacts and contributes to our understanding of community level response to invasion. The ability to predict how species will respond to invasion has multiple applications. It can inform restoration by providing a guideline for evaluating the ability for native species to tolerate the conditions created by non-native species. It can also assist in setting conservation priorities by identifying which species might be susceptible to the invasion of non-native species, especially those with the potential to alter the physical structure and resource availability in the regions they invade. In our study, we used traits as indicators for how species were responding to changes in their environment; however, they can also be used to predict how ecosystem services may change following a perturbation, like the invasion and rise to dominance of a non-native species (Funk et al. 2008; Suding et al. 2008). One important future direction is scaling our community level findings to ecosystem processes to evaluate the effect of invasion on grassland function.

Acknowledgements

We thank Viviane Vincent and Jono Wilson, for their help in the field and laboratory, Karen Stahlheber and Jono Wilson for contributing ideas and statistical advice, Susan Harrison and Marcel Rejmanek for comments that improved the manuscript, and the Mildred Mathias UC Reserve fellowship and the California Native Plant Society for contributing funding for travel and equipment.

Data accessibility

Data deposited in the Dryad repository: http://doi.org/10.5061/dryad.50hd2 (Molinari & D' Antonio 2014).

Ancillary