Introduction

When an ecosystem is invaded by a shrub, tree or vine the structural changes (e.g. Mimosa pigra (Gordon 1998)) or homogenising effects (e.g. Chrysanthemoides monilifera (French et al. 2008)) on the local vegetation might be more visually obvious than if the invader is a annual grass or forb. Grassy ecosystems can have great spatial and temporal heterogeneity at a fine scale (sub-meter), and so the interaction between exotic and native plants may occur at a similarly fine spatial scale (Dillemuth et al. 2009; Theoharides and Dukes 2007). Exotic plant invasions which alter the ground layer vegetation structure and lead to loss of heterogeneity could have impacts on many aspects of grassy ecosystems, and this is the focus of this study.

Heterogeneity in the ground layer vegetation can be created in several ways (Fig. 1). Having large gaps in the vegetation can create one type of heterogeneity (Fig. 1b), even if these gaps are spaced evenly. Other ways to achieve heterogeneity are to have a range of gap sizes (Fig. 1c) and a range of plant sizes (Fig. 1d). Fine scale spatial heterogeneity and gaps within grassy ecosystems can be important in maintaining plant (DiVittorio et al. 2007) and animal diversity (Allan and Southgate 2002). For example, the spatial heterogeneity and existence of gaps in Spinifex grasslands (Triodia species) encourages fine-scale partitioning of space by animals. This is thought to be one of the reasons for the large richness of reptiles in these grasslands, as they can forage in the tussocks, at the edges and spaces between them (Cogger 1984). Ground layer complexity has also been shown to affect ant species richness (Lassau and Hochuli 2004), with the availability of structurally simple spaces important for thermophilic ant species in grassy ecosystems (Lindsay and Cunningham 2009).

Fig. 1
figure 1

Simplified depictions of different ground layers in grassy woodlands, with either homogeneous or heterogeneous spatial arrangements. a homogeneous (minimal variation in basal area and no gaps), b heterogeneous type 1 (minimal variation in basal area and large evenly spaced gaps), c heterogeneous type 2 (minimal variation in basal area but variation in gap size) and d Heterogeneous type 3 (Variation in basal area and gap size)

Full size image

Canopy gaps in grasslands can be important for the germination and survival of forbs (Morgan 1998), with both the size of gaps (Silvertown and Smith 1989) and the arrangement of gaps (Platt and Weis 1977) influencing plant recruitment. Small scale variation in plant cover may also allow more plant species to coexist (Silvertown 1981). Such heterogeneity may increase the success of a plant invader, but it can also reduce impacts on native plant species, as it promotes coexistence mechanisms not possible in homogeneous environments (DiVittorio et al. 2007; Melbourne et al. 2007).

The loss of gaps from grassy ecosystems could have consequences for the movement and capture of organic matter. Perennial grass tussocks are very effective in trapping soil particles, litter fragments and seeds carried by the wind (Ludwig et al. 2005). Water and nutrients captured by perennial vegetation can increase patch growth, and this will maintain or enhance the capacity of the patch to obstruct future overland flows. This positive feedback can help to maintain patchy vegetation in low rainfall areas (Ludwig et al. 2005). The replacement of perennial grasses by annual grasses could alter the capacity of the ground layer to capture resources all year round, and could lead to resources being trapped within smaller areas during the annual grasses peak growth phase (Tongway and Ludwig 1997).

The ground layer of intact grassy woodlands in south-eastern Australia is dominated by perennial grasses. Most of the floristic diversity in these woodlands is in the understorey, with the inter-grass spaces occupied by a range of herbaceous species (Prober and Thiele 2004). A variety of grass structural types occur in grassy woodlands including tall tussocks, caespitose and decumbent perennials. Within a given site, usually only a few grass species dominate, with most species having a low cover. The variety of plant life forms and mixture of naturally common and rare species usually creates structurally heterogeneous ground-layer.

Many species of exotic annual grass (e.g. Bromus diandrus, Hordeum leporinum, Lolium rigidum) are known to invade grassy eucalypt woodlands. Generally several exotic species invade woodlands together, such that an invaded understorey is rarely monospecific. These annual grasses are however, similar in structure and leaf traits and have a lower C:N ratio and leaf mass area then the perennial grasses they replace (McIntyre and Lavorel 2007). Complete replacement of the woodland understorey by exotic plant species can occur, and exotic annual grass dominance generally persists once exogenous disturbances are removed (Lindsay and Cunningham 2011).

For many invaded ecosystems there is still uncertainty as to whether invasive species are ‘drivers’ of community change or are ‘passengers’ of changes in environmental condition. (MacDougall and Turkington 2005). The driver model is when invasive species limit or exclude native species by competition, whereas the passenger model is when the invasive species is less affected by disturbance or recruitment barriers. The ‘driver’ and ‘passenger’ models are not always exclusive, and can interact (MacDougall and Turkington 2005). There has been considerable progress over the past 10 years in understanding the drivers of annual grass invasion in woodland and grassland remnants (e.g. Lenz and Facelli 2006; Lindsay and Cunningham 2011; McIntyre and Lavorel 2007; Prober et al. 2002b). These findings all suggest that the ‘passenger’ model is the underlying cause of the initial grass invasion, with invasion mainly occurring in woodlands disturbed by agricultural practices (e.g. livestock grazing, fragmentation) or that are in close proximity to agricultural land (Prober et al. 2002b). However, it is not known if exotic annual grasses are also the drivers of ecosystem change, and further impact the ecosystem once they have become established.

Exotic plant invasions may make the vegetation structure more homogeneous (Gentle and Duggin 1997; Gordon 1998; Parker 2000), but this is rarely quantified, especially for grass dominated systems (Dillemuth et al. 2009). In this paper we examine the effects of exotic annual plant invasion on the woodland ground-layer. We aimed to determine if annual grass invasion in woodlands created a more homogeneous (‘lawn-like’) ground layer structure (Fig. 1a). We hypothesised that native uninvaded woodlands naturally have a heterogeneous understorey; i.e. that there is a range of plant sizes and they occur in both clumped and patchy distributions (Fig. 1d). We addressed the following questions:

  1. 1.

    Is the interplant distance and basal plant area smaller for exotic grasses than native plants?

  2. 2.

    Does the type of neighbouring grass affect the spacing and size of grasses?

  3. 3.

    Is there less variation in interplant distance and basal plant area for exotic than native plants?

  4. 4.

    Do weed invaded woodlands have a greater plant density and does native plant density decrease with increasing total plant density?

  5. 5.

    Does total plant area increase with plant density?

We conducted surveys in the ground-layer of eucalypt dominated grassy woodlands that varied in exotic plant cover, from sites with no exotic species through to sites dominated by exotics, with the contrast between the dominant woodland understory plant group, the native perennial grasses and exotic annual grasses the main focus.

Materials and methods

Field sites and vegetation

We surveyed 22 grassy eucalypt woodland sites in spring (September–November) 2008. The sites were located across a 170 km line from 35.250S, 149.433°E to 34.177S, 148.217°E on the southern tablelands and south west slopes of NSW Australia. Mean annual rainfall in this region is 554–725 mm, but it was below average in the year of the surveys (Australian Bureau of Meteorology). Tree canopy cover ranged from 10 to 24% and this was created by one to four Eucalyptus species at each site, with the most common species being E. blakelyi, E. melliodora, E. albens, E. goniocalyx, E. mannifera, E. machrorhyncha, E. nortonii and E. racemosa. There was no midstorey cover at most sites (n = 17 sites), and the maximum midstorey cover was only 2.2%. Plant nomenclature follows Harden (1992–2007), with updates from PlantNET (1999–2009). All sites had a history of intermittent grazing by livestock (>100 years) including periods of set-stock grazing. Over at least the past 5 years the sites had experienced much lower levels of grazing by either sheep, native (e.g. Macropus giganteus) and/or feral herbivores (e.g. Oryctolagus cuniculus). None of the study sites had been burnt for at least 10 years.

The native understorey vegetation, where present, was dominated by perennial C3 and C4 species. The common native grasses were; Aristida ramosa, Austrostipa scabra, Austrostipa bigeniculata, Austrodanthonia sp., Elymus scaber, Joycea pallida, Microlaena stipoides, Poa sieberiana, Sporobolus creber and Themeda australis. The common rushes and sedges were Lomandra bracteata L. filiformis, L. mulitflora, Luzula densiflora, and Juncus spp. There was a high diversity of native forbs, all in low abundance. All shrubs (Family Epacridaceae) were prostrate or short (<0.5 m high). Prober et al. (2002a) and Prober and Thiele (2004) provide further information about these woodland communities.

The most common exotic grasses were cool season (C3) annuals. These were Aira elegantissima, Avena fatua, Briza major, Briza minor, Bromus diandrus, Bromus hordeaecus, Hordeum leporinum, Lolium rigidum, Lolium spp., Poa annua, Vulpia bromoides and Vulpia myuros. The common exotic forbs were; Acetosella vulgaris, Arctotheca calendula, Cerastium glomeratum, Echium plantagineum, Erodium spp., Hypochaeris radicata, Hypochaeris glabra, Petrorhagia nanteuilii, Trifolium glomeratum, T. campestre, and T. augustifolium.

Sampling design

We used the wandering quarter technique to measure plant density and interplant distance. Interplant distance was used as a measure of gap size. The wandering quarter is a plot-less method for estimating relative density (Catana 1963), and is suitable for use in grass dominated systems (Becker and Crockett 1973). In each site we established a 50 m quadrat with one side parallel to the greatest slope at the site. Inclines did not exceed 15º in any site. We conducted four wandering quarter transects in each quadrat, each 15 m long. The point of origin for each wandering quarter was selected by structured random sampling such that the four transects were evenly spaced across the slope of the quadrat (10 m apart) but with starting points at 0, 15, 25, 35 m down slope. From each starting point we measured the distance to base of the nearest plant within a 90º arc centred down slope. This means that progress was often in a ‘zigzag’ fashion when the vegetation was at low density. We continued to do this until a 15 m length was covered. The starting points were sufficiently spaced (10 m), and the vegetation dense enough so that no plant was measured twice. We recorded the identity of each plant to species where possible and measured the basal area. Width and breadth were measured at the butt of the plant close to ground level. We only measured plants in the ground layer, most of which were grasses. The prostrate shrubs were highly compact, and for these we measured the area in contact with the ground. We did not measure plant height, as all sites were grazed by native and introduced herbivores. A total of 1,320 m in length was surveyed over the 22 sites. We also noted the interplant ground cover type as leaf litter, rock, bare soil or cryptogam.

Data treatment

We categorised plants into the following groups: native grasses (G), native forbs (F), native shrubs (S), native rushes and sedges (R), exotic annual grasses (EAG) and exotic forbs (EF). There were no exotic perennial grasses in our sites. Shrubs were the rarest plant group, occurring at a very low density (1 plant per 250 m for sites where present) and only on the native vegetation dominated transects. Due to the low number of data points we could not compare parameters for the shrubs across the different dominant vegetation types, and so shrubs were excluded from the main analyses. Shrubs are often lost with overgrazing and disturbance (Prober and Thiele 2004).

We calculated the mean and inter-quintile range (80% quantile–20% quantile) for each transect for the interplant distance and basal area for each plant group. We calculated the quintile range as we were also interested in the spread or variability in the data and this measure provides a descriptor of spread that is little influenced by outliers (Quinn and Keough 2002). We also calculated the plant density per meter for each plant group on a transect basis.

We hypothesised that the dominant plant type or the nearest neighbour could affect a plant’s basal area and spacing. To examine this we characterised each transect by the dominant plant type into one of three groups, native (<33% exotic vegetation cover), mixed (33–66% exotic) or exotic (>66% exotic). To investigate effects of dominant plant type on interplant distance and basal area we initially examined data on a transect basis. After establishing that dominant plant type was a significant influence in some analyses, we subsequently investigated the effect of nearest neighbour on the interplant distance and basal area with a second set of analyses only including the native and exotic grasses (i.e. shrubs, ruches, sedges and forbs were excluded). For these analyses we only included data for grasses where the two nearest neighbouring plants (upslope and down-slope) were also grasses from the same group as each other (i.e. G–G or EAG–EAG). This enabled us to compare the impacts of the two extreme cases, either EAGs neighbouring up and down slope or native grasses up and down slope. We randomly excluded some of the exotic grass data so that there were the same number of data points as for the native grasses (total n = 3,423). This was done with the Generate Random Sample procedure in GenStat.

Statistical analysis

All statistical analyses were carried out in GenStat version 12.1.0 (Payne et al. 2009). We tested for differences in the interplant distance and basal plant area between the plant groups by a three factor residual maximum likelihood (REML) analysis. The fixed factors were plant group (five levels) and transect dominant plant type (three levels), whereas site was treated as a random factor. These analyses were performed with both the transect mean and inter-quintile range for each plant group. All data were square root transformed to improve homogeneity of variance and normality. For post-hoc contrasts we compared predicted means from the REML model with the least significant difference (LSD) at α = 0.05. We examined models with transect as a factor, but this was not significant in any analysis nor were there any interactions between transect and the other factors, therefore it was dropped from the final models. We used REML instead of analysis of variance as not all plant groups occurred at all sites (five sites had no exotic plants, and one site had no native plants) thereby creating an unbalanced design.

REML was also used to investigate the impacts of neighbouring grass type on the interplant distance and basal area for the native and exotic grasses. Focal plant type (two levels: EAG or G) and neighbour type (two levels: EAG or G) were treated as fixed factors, and site was a random factor. This analysis could only be performed on the raw data (square root transformed) and not the transect average data, such that the inter-quintile range could not be investigated.

We conducted a REML analysis to compare the mean density of the different plant groups. We did not include transect dominant plant type as a factor in this analysis, because plant densities were used to define the dominant plant types, creating a circularity. We also used REML to compare the total plant density for each dominant plant type.

For the regression analyses we placed data into only four plant groups: exotic annual grass (EAG), exotic forb (EF), native grass (G) and other native (ON). This last category included the forbs, lilies, sedges, rushes and shrubs, which were grouped together due to their low density. We used the General Linear Model (Normal, logit) in GenStat and all data were square root transformed. Again, transect was not a significant factor (P > 0.869), and there was no significant higher order interaction with the transect factor in any of the regressions, so it was not included in the final models. All regression analyses were conducted with data calculated on a transect level, unless otherwise stated. We performed two sets of regressions. First, to investigate if the density of each plant group (response variable) responded differently to increases in total plant density (predictor variable), we conducted a regression with plant group as a grouping factor. We examined differences in the regression lines (slopes and intercepts) for each group to that of both the native and exotic grasses. Second, we investigated if there was a relationship between total basal area and total plant density. There was no clear relationship between these two variables so we looked at the relationship between the total basal plant area and exotic annual grass density and native grass density separately to see if these two plant groups were behaving in different ways.

Results

We surveyed 11,964 plants across all the sites. For the five sites where exotic plants were absent, perennial grasses were the dominant plant type with a density of 2.40 ± 1.9 plants m−1 (Mean ± 1 SD). Rushes and sedges were at a similar density to the forbs and lilies, at 0.76 ± 0.64 and 0.71 ± 0.81 plants m−1. Litter was the most common interplant ground cover type (78 ± 23%), followed by rock (18 ± 22%) and cryptogam (moss or lichen) (4.2 ± 11%).

Density relationships between plant groups

EAGs occurred at a higher density than any other plant group (Mean 5.36 ± 4.7 SD plants m−1) (Plant group Wald = 212.1, d.f. = 4, 12, P < 0.001). Native grasses occurred at a higher density (2.18 ± 1.9 plants m−1) than native rushes (0.379 ± 0.59 plants m−1), forbs (0.595 ± 0.79 plants m−1) and exotic forbs (0.573 ± 0.97 plants m−1). EAGs reached the highest density, 17.5 plants m−1, with native grasses only reaching 8.07 plants m−1.

The exotic dominated transects had a similar mean plant density (11.8 ± 2.5 plants m−1) to the mixed vegetation transects (10.6 ± 3.7 plants m−1), with both having a higher density than the native dominated transects (5.63 ± 2.4 plants m−1) (Wald = 68.51, d.f. = 2, 70, P < 0.001).

For each plant group we detected a significant (P ≤ 0.048) relationship between the density of the group and the total plant density (R2 = 57.9, F7,344 = 70.0, P < 0.001) (Table 1). EAGs and EFs both had positive correlations with total plant density (Fig. 2), whereas the density of native grasses and non-grasses were negatively correlated with total plant density (Table 1). We examined regressions for each of the group versus total density relationships (e.g. Fig. 2) with and without zeros to determine the reliability of our conclusions, and found that in each case the P value remained <0.001. The slopes and intercepts of the regression lines for native grass and non-grass density differed to that of the exotic grasses (P < 0.001) (Table 1).

Table 1 Summary of the linear regression analysis for total plant density versus the density of each plant group
Full size table
Fig. 2
figure 2

The average a exotic annual grass density and b exotic forb density in comparison to the total plant density. The zeroes were not excluded as these graphs are part of a regression model involving all plant groups as summarised in Table 1. The 95% confidence intervals are also shown. All data were square root transformed

Full size image

The results of the density regression were mirrored in the correlations between plant groups. There was a negative correlation between native non-grass density with each of exotic forb density (r = −0.326, P < 0.002) and exotic annual grass density (r = −0.527, P < 0.001), and a negative correlation between native grass density and exotic annual grass density (r = −0.595, P < 0.001).

Interplant distance

Interplant distance was greater between native grasses and neighbours than for any other plant type (Table 2; Fig. 3). There was also a greater range in the interplant distances for the native grass and rushes than the other groups (Table 2). Post hoc analysis revealed that native forbs and rushes also had a greater mean interplant distance than the EAGs and EFs (Table 2; Fig. 3).

Table 2 Summary of the REML analysis for comparison of interplant distance and basal plant area (mean and inter-quintile range) between the plant groups (native grasses, exotic grasses, native forbs, exotic forbs, native rushes) and the dominant vegetation of the transect (Native, mixed, exotic). Bold indicates P ≤ 0.05
Full size table
Fig. 3
figure 3

The predicted means for interplant distance (cm) for each plant group in comparison to the dominant plant type of the transect (native, mixed or exotic). Means were calculated on a transect basis, and the dominant plant type was also assigned on a transect basis. The least significant difference (LSD) from the REML analysis and the back transformed means are also indicated

Full size image

All plant groups were separated by greater distances when in native dominated transects compared to the exotic or mixed vegetation transects (Table 2; Fig. 3). Native grasses in native dominated transects (24 ± 14 cm) were spaced more than twice as far apart as the EAGs in exotic dominated transects (8.1 ± 1.8 cm). When the grass interplant spacing were examined in more detail, the REML analysis showed both the native and exotic grasses were spaced further apart when they had native grasses as neighbours (upslope and down-slope) instead of exotic grasses (Neighbour: Wald = 680.3 d.f. = 1, 3410 P < 0.001).

Plant basal area

Native grasses had the largest basal area and had a greater range in basal areas than all other plant groups (Table 2; Fig. 4). The rushes also had a larger basal area (Fig. 4) and more variation in basal area than the exotic grasses and exotic forbs (Table 2). The EAGs and EFs were of similar size. There was no effect of transect dominant plant type on basal plant area, and plant area and interplant distance were only weakly correlated (r = 0.095, P < 0.001).

Fig. 4
figure 4

The predicted means for basal plant area (cm2) for each plant group. Dominant plant type factor was not significant and is not shown. The least significant difference (LSD) from the REML analysis and the back transformed means are also indicated

Full size image

When we looked at the grasses in more detail the REML analysis showed the native grasses had a greater basal area when they had EAGs neighbouring either side instead of native grasses (Grass type × neighbour: Wald = 1.37, d.f. = 1, 3414, P ≤ 0.02). EAG basal area did not vary with the type of neighbouring grass.

Basal plant area and density relationships

Total basal plant area increased as native grass density increased (R2 = 18.0, F1,86 = 19.9, P < 0.001) and the native dominated transects had a greater total basal plant area than the exotic dominated transects (0.102 vs. 0.069 m2) (Dominant plant type, Wald = 6.14, d.f. = 2, 463, P = 0.047). However, the total plant area tended to decrease as exotic annual grass density increased (R2 = 4.5, P = 0.027, F1,86 = 5.08). These two opposing effects tended to cancel each other out, so that there was no relationship, either linear or non-linear, between total plant density and total plant area (P ≥ 0.153). The plant area inter-quintile range decreased as the total plant density increased (R2 = 26.6, F1,86 = 31.7, P < 0.001).

Discussion

There was lower spatial heterogeneity in the ground-layer of grassy woodlands heavily invaded by exotic annual grasses compared to woodlands with little to no exotic grass cover. This diminished heterogeneity is created both by the tendency for exotic annual grasses to be smaller and more closely spaced, and further because there is less variation in basal area and spacing. Even when exotic forbs, such as Acetosella vulgaris, were also present the horizontal structural diversity was still low in exotic dominated sites. This was because the exotic forbs were arranged in a similar way to the exotic grasses, generally being smaller and closer together, and with less variation than the native forbs. These kinds of changes in structural heterogeneity in the presence of invasive grass species have rarely been quantified (Dillemuth et al. 2009), particularly with regards to the magnitude or scale of the change.

As expected, there was substantial variation in the horizontal structure in the ground-layer of uninvaded woodlands. The native vegetation was dominated by long-lived perennials. With basal area generally increasing with age (depending on resources and grazing pressure), this creates the potential for a range of age and size classes to exist at one point in time. In contrast, the EAGs generally live for less than 6 months in south-eastern Australia (late winter through to early summer) (Jessop et al. 2006). Even though they tend to allocate more resources to shoot growth and reproduction and less to root growth than the native grasses, their size is constrained by their short life.

Unexpectedly, total basal area decreased as plant density increased. This was because high plant densities only occurred in sites dominated by EAGs, and the EAGs had a smaller and less variable basal area than the native grasses, rushes and sedges. The maximum density reached would be affected by numerous factors including resource availability, herbivory, ground disturbance and space. There is a very large viable soil seed bank for many of these exotic annual grasses (Lindsay unpublished data), so seed supply is unlikely to limit density. The surveys were, however, conducted when rainfall had been below average for the past 4 years, so any increase in rainfall is likely to make the ground layer in the invaded sites more homogeneous.

Invaded woodlands had a greater overall plant density and a lower native plant density. Due to the heterogeneity and existence of gaps in the native woodlands it is possible that the EAGs are filling in some of these gaps. However, the reduction or complete absence of native ground-layer vegetation from invaded sites suggests the EAGs are more than ‘gap fillers’ in the grassy ecosystems we examined, and that they are also replacing or displacing the native plants. Replacement might occur if native plants are overgrazed and unable to set seed, whereas displacement would occur if the EAGs can outcompete the native grasses. EAGs are more likely to outcompete the native grasses when soil nutrients are elevated (Groves et al. 2003), which is often the case for woodlands which have been grazed by livestock in the past (Lindsay et al. 2010).

Just as the invasive Bromus inermis (smooth brome) is detrimental to the patch dynamics of Spartina pectinata (prairie cordgrass) grasslands in North Dakota (Dillemuth et al. 2009), it is likely this assemblage of EAGs are detrimental for the recruitment, survival, growth of native plant species. The uninvaded woodlands contained a range of gap sizes, and as is typical, grasses only occupied 30–50% of the ground cover (Lunt et al. 1998). Gaps provide spaces for recruitment and both gap size and gap distribution influence the average distance between individuals in consecutive generations (Bergelson et al. 1993). The lack of medium (>10 cm length) to large gaps (>30 cm length) in the exotic dominated woodlands is likely to determine the suitability of techniques for successful re-introduction of native grasses for ecological restoration. Suitable micro-sites may need to be created for some species by herbicide application, burning or tilling (Clarke and Davison 2004).

Competition for light can be an important mechanism driving plant diversity loss when productivity increases after soil nutrient enrichment (Hautier et al. 2009). In our system EAG invasion is rarely successful in woodlands which do not have a history of disturbance, with EAG invasion also associated with increased soil nitrate or phosphorous (Dorrough et al. 2006; Lindsay and Cunningham 2011; Prober et al. 2002b). The native understorey flora generally responds negatively to increased nitrate and phosphorous, with only a few species increasing in cover when nitrogen is elevated (e.g. Microlaena and Austrodanthonia spp.) (Lodge and Whalley 1985). It is possible that the nutrient enriched EAG dominated sites could have higher productivity during late winter and spring than the uninvaded sites. This combined with the small gap size could reduce the number of patches where the ground receives adequate light for native seedlings and small forbs to survive following EAG invasion.

The findings presented here suggest that EAGs are drivers of community change, and the findings of previous studies suggest that the EAGs are passengers of agricultural disturbance. We do not know if the passenger or driver model is predominately responsible for the EAG dominance, or if both models are of similar importance in invaded grassy woodlands. MacDougall and Turkington (2005) concluded that the impacts of invasive grasses on community structure in an oak woodland were more dominated by noninteractive factors then competition, such that the passenger model better explained exotic grass dominance. However, they also found that the driver model played a significant role in shaping the community structure, with the exotic grasses also having suppressive and facilitative effects.

The reduction in spatial heterogeneity we quantified in the EAG invaded woodlands was at a small scale (<1 m). This work does, however, have implications at larger scales, with cool season annual grass invasion occurring at large scales (<1 ha) in grassy eucalypt woodlands, Australian temperate grasslands (Lenz and Facelli 2006) and various grassland, oak woodland and grass-shrub communities in North America (e.g. Belnap and Phillips 2001; Holmes and Rice 1996; Mack 1981).

While individually these exotic annual grass species may pose little threat to the woodland ground-layer, our study shows that as an assemblage they have a transforming effect on the habitat structure. The woodlands we investigated are a part of an endangered ecological community recognised by the Australian Government (Environment Protection and Biodiversity Conservation Act 1999), with less than 5% of the original cover remaining and most in poor condition. Control and management of these EAGs in native vegetation has been limited. This work illustrates that once EAG have invaded and naturalised, they can alter the structure of grassy woodlands in a way that needs to be considered when managing for conservation or biodiversity outcomes. Methods for re-introduction of native grass into invaded woodlands and how to shift the competitive balance to native grasses in invaded woodlands are being trialled (Cole et al. 2004; Lindsay and Cunningham 2011; Prober and Lunt 2009), but are yet to be tested on a site scale.