- Top of page
- Materials and methods
- Supporting Information
Riverine and floodplain ecosystems possess many features that promote invasion by exotic plant species, including high perimeter: area ratios, propagule transport in flowing waters, creation of bare ground by floods, productive soils and temporally variable resource levels (Scott, Auble & Friedman 1997; Davis, Grime & Thompson 2000; Richardson et al. 2007). Anthropogenic processes, such as flow regulation and livestock grazing, may further promote invasions by altering natural resource levels and disturbance regimes (Jansen & Robertson 2001; Baron et al. 2002; Tockner & Stanford 2002). Human settlements and activities, which tend to be concentrated around rivers, may also be a source of propagules of invasive species (Tockner & Stanford 2002).
Attempts to restore invaded floodplains can be impeded by many factors. Flows are often regulated and contested (Baron et al. 2002), and degraded processes may take long periods to recover (Tockner & Stanford 2002). Exotics may remain abundant after degrading processes are removed and may alter abiotic conditions and biotic interactions in ways that prevent the recovery of native species (Levine et al. 2003; Prober, Lunt & Morgan 2009). Thus, floodplain managers face major challenges to (i) identify how a wide range of processes influence invasion, (ii) distinguish between reversible and irreversible processes and (iii) control and manipulate degrading processes that operate over large areas and long time-spans.
Many floodplains are dominated by a single functional group of exotic plants: annual grasses and forbs, many of Mediterranean origin (e.g. Avena, Bromus, Hordeum, Trifolium, Medicago, Arctotheca, Echium and Hypochaeris species). These taxa are also widespread and abundant in adjacent, dryland ecosystems in many regions of North America and Australia (Mack et al. 1989; Prober & Thiele 1995; Bartolome, Jackson & Allen-Diaz 2009). Two contrasting theories have been proposed to explain the successful invasion of floodplains by exotic annuals. Traditionally, invasion has been attributed to heavy grazing by introduced livestock, as in dryland ecosystems. In ecosystems with limited exposure to large grazing animals over evolutionary periods, intensive livestock grazing after European settlement caused dominant native perennial plants to be replaced by exotic annuals (Milchunas, Sala & Lauenroth 1988; Cingolani, Noy-Meir & Diaz 2005). In productive areas, this change is difficult to reverse even if livestock are removed, as fast-growing annuals capture soil resources at the start of each growing season and prevent establishment of slower-growing native perennials (McLendon & Redente 1992; Prober et al. 2005). This transition has been characterized as a non-equilibrium dynamic, with alternative stable states dominated by either native perennials or exotic annuals (Bartolome, Jackson & Allen-Diaz 2009; Prober, Lunt & Morgan 2009).
Recently, an alternative explanation for invasion has been proposed, based on floodplain surveys conducted during extended drought periods. This ‘terrestrialization’ hypothesis (Catford et al. 2011) posits that annual, terrestrial exotic plants have invaded floodplains following a reduction in the frequency, magnitude and duration of floods (Stokes, Ward & Colloff 2010; Catford et al. 2011). Catford et al. (2011) modelled past and present river flows and found that exotic cover was greatest in intermittent wetlands that had experienced the greatest reduction in flooding.
To disentangle the impacts of livestock grazing and flooding on plant invasions, the effects of historic and current grazing regimes must first be distinguished (Borman 2005; Lunt et al. 2007a). Current grazing regimes bear little relationship to historic regimes that occurred before current land uses, tenures and townships existed. Many floodplains were intensively grazed by livestock long before rivers were regulated. Consequently, the degree to which historic grazing practices may have predisposed floodplains to invasion following river regulation cannot be determined.
This caveat aside, the grazing and terrestrialization models may be conceptualized as being relevant at opposing ends of the topographic sequence, with the grazing model applying primarily at higher elevations and the terrestrialization model in low-lying areas. However, flood patterns vary greatly over time, and floodplains contain fine-scale topographic mosaics, with subtle changes in inundation across small elevational changes and short distances (Chesterfield 1986). Consequently, it is difficult to disentangle the influence of the two processes in many parts of the river floodplain.
Regardless of the processes that promoted initial invasion, a key question for conservation management is to what extent do current grazing and flooding regimes, and their interactions, affect established populations of exotic species. Can exotic species be controlled by manipulating either or both of these processes? Theory suggests that exotic annuals are likely to be more easily controlled by manipulating flooding than livestock grazing. In productive ecosystems with a short evolutionary history of grazing, the replacement of native perennials by exotic annuals caused by livestock is widely viewed as an irreversible (or non-equilibrium) dynamic (Cingolani, Noy-Meir & Diaz 2005; Standish, Cramer & Hobbs 2008). Recovery of native species is constrained by many factors, including the loss of native species and their propagules, rapid biomass accumulation by fast-growing exotics when disturbances are removed and altered rates of soil water and nutrient cycling, which exotic annuals can control (Standish, Cramer & Hobbs 2008; Prober, Lunt & Morgan 2009). Indeed, grazing exclusion led to minor changes in the composition of exotic and native plants in a 12-year experiment in a rarely flooded floodplain (Lunt et al. 2007b).
By contrast, flooding changes the availability of atmospheric and soil resources for plants, leading to changes in plant composition from dryland to wetland species. Wetland plants commonly possess large soil seed banks, enabling rapid recovery following inundation (Brock et al. 2003). Thus, flooding–drying cycles may be expected to result in predictable and reversible transitions between wetland and dryland plant communities as water levels change (Brock et al. 2003), perhaps with little biotic interaction in many years between either functional group. Interactions between flooding and grazing regimes may also affect exotic and native species. For example, spatial variability in flood patterns can influence animal movement patterns (by promoting food availability), controlling the distribution and intensity of grazing impacts (Smith et al. 1992). If floodwaters are shallow, then grazing intensity may increase during floods and after flood recession, with direct and indirect impacts on both wetland and dryland plants.
The Murray-Darling Basin is eastern Australia's largest water catchment, supplying cities, irrigated agriculture and natural ecosystems. Water supplies are heavily utilized, with dams able to store more than the average annual runoff and almost 90% of divertible water is extracted (Kingsford 2000). Thus, environmental allocations are vigorously contested (Murray-Darling Basin Authority 2010). This situation is expected to worsen with climate change. Consequently, it is imperative that environmental allocations are well targeted. While flooding has been suggested as a practical method to control exotic annuals (Stokes, Ward & Colloff 2010; Catford et al. 2011), no experimental data are available on how floods in different seasons affect exotics. Exotic annuals emerge in autumn, die after seed fall in late spring and persist over summer in the soil seed bank (Rossiter 1966). Consequently, floods in summer (when no standing plants are present) may have less impact on populations than floods in autumn or spring, after seeds have germinated.
In this study, we assess the effects of flooding and livestock grazing on the cover of exotic annuals in a floodplain forest in south-east Australia. We use data from an 8-year flooding and grazing experiment to ask (i) does flooding in different seasons have different impacts on exotic annual species, and (ii) how do interactions between flooding and livestock grazing affect exotic annuals. We compare responses of exotic annuals against native species to illustrate treatment effects on the entire floodplain community. This study provides the first experimental investigation of the effects of floods in different seasons, with and without livestock grazing, on established populations of exotic annuals.
Materials and methods
- Top of page
- Materials and methods
- Supporting Information
The study was conducted in the former Gulpa Island State Forest (35°45′S, 144°54′E), now Murray Valley National Park, on the Barmah-Millewa floodplain. The climate is semi-arid, with an annual average of 460 mm rainfall and 1530 mm evaporation (Bacon et al. 1993). The forest is dominated by Eucalyptus camaldulensis (height 25–40 m; Cunningham et al. 1981), with a herbaceous ground cover (mainly exotic annual and native perennial grasses and herbs). Forest growth is dependent on regular flooding in this climate. Prior to river regulation, floods occurred mainly in spring (August–November) and the floodplain was inundated in most years (Robertson, Bacon & Heagney 2001). River regulation and water extraction have resulted in fewer floods in spring and more short floods in summer (Robertson, Bacon & Heagney 2001).
The floodplain is dissected by shallow stream beds (‘flood runners’), ca. 10–20 m wide and 0·5 m deep, that support wetland vegetation when flooded and terrestrial plants, including exotic annuals, when dry. In this experiment, flood runners were artificially flooded by pumping and impounding water in different seasons, from 1990 to 1998. In 1990, one of four flooding regimes was allocated to each of four separate flood runners in three replicate sub-regions in Gulpa Island State Forest – no flooding, summer floods, spring floods and summer + spring floods – giving 12 flood runners in total. All 12 flood runners were flooded in spring 1990 to allow plots to be laid out in relation to maximum flood levels. Spring floods extended from early September–October (when annual exotic species were mostly in their active growth phase, prior to flowering and setting seed), while summer floods occurred in January–February (after most annual exotic species had finished seeding). Artificial flood waters were maintained at a constant level for ca. 4 weeks from 1991 to 1994, and for ca. 8 weeks in the following experimental years, and water levels were left to recede naturally (Robertson, Bacon & Heagney 2001). Flood waters were <0·9 m deep at the deepest point in each flood runner (Bacon et al. 1993). Despite attempts to control flood regimes, natural floods inundated all flood runners for short periods of time in spring 1992 and 1996 and a longer period in 1993.
In 1990, before experimental flood regimes were imposed, a pair of plots, which were subsequently randomly allocated to grazed and non-grazed treatments, was established on each flood runner. Each plot was 10 m wide and c. 25 m long and ran upslope from the bottom of the flood runner to the adjacent unflooded area. Three parallel transects were placed in each plot, and 1-m2 quadrats were permanently marked along each transect. In this study, we analyse data from low-lying quadrats that were subject to experimental flood regimes. A companion study documented the effects of grazing on unflooded quadrats further upslope (Lunt et al. 2007b). Three, 1-m2 quadrats subject to flooding were sampled from the flood runners along each transect in each plot, giving nine quadrats per plot and a total of 216 quadrats each year.
In autumn 1991, grazing exclusion plots were protected by 1·2-m high fences, which excluded stock and rabbits, and reduced (but did not eliminate) grazing by grey kangaroos (Macropus species), which were present at low densities. Small numbers of rabbits accessed some plots in the later stages of the experiment. Thus, this experiment only examined the effects of livestock grazing, with a low background level of grazing by rabbits and kangaroos occurring across all plots. Grazed areas were stocked with cattle at a relatively low stocking rate of approximately 0·3–0·6 DSE (dry sheep equivalents) per hectare per annum. The three sub-regions were separated by at least several hundred metres, and the grazed study area occupied 5960 ha in total.
Vegetation was surveyed in late spring to early summer (October–December) in 5 years: 1990 (prior to erection of the fences and initiation of flooding regimes), 1991, 1992, 1995 and 1998. Spring floods had receded before sampling began, except in 1995 when water up to 20 cm deep remained in some quadrats. Within each quadrat, the cover of each vascular plant species was estimated to the nearest 5% (if >10% cover) or nearest 1% (if <10% cover). By sampling in spring, we documented the effects of season of flooding on terrestrial vegetation in the spring following flood events, rather than at the time of flooding (c.f. Robertson, Bacon & Heagney 2001). Species names follow the study by Harden (2000).
Herbaceous species were classified into functional groups according to origin (native vs. exotic), life-form (annual vs. perennial) and habitat (terrestrial, mudflat vs. aquatic) based on Cunningham et al. (1981), Harden (2000) and field observations. Data from the nine quadrats in each plot were averaged before analysis, giving one value per plot. Data were analysed using PERMANOVA in PRIMER (Clarke & Gorley 2006; Anderson, Gorley & Clarke 2008). Details of the analysis are provided in the Supporting Information. In brief, a split-plot repeated measures design (Quinn & Keough 2002) was used to examine the effects of year, flood regime, grazing and sub-region, where year was the repeated factor, sub-region was a random factor and grazing was applied as a within-plot treatment, while flooding regimes were applied between plots across all three sub-regions.
The experimental design includes a pre-treatment sampling time (1990) and four post-treatment sampling times (1991, 1992, 1995 and 1998). In 1990, all plots were flooded in spring and grazed. Subsequently, the grazing and flooding treatments were applied, although in the spring of 1992 all plots were inundated by a natural flood. Thus, the effects of flooding and grazing are revealed by examining the flooding × year interaction and the grazing × year interaction (Quinn & Keough 2002). This approach allows for large differences between plots pre-treatment and also enables examination of trends over time. As there was large inter-annual variability in most measures, planned comparisons of pre- and post-treatment data did not occur. Instead, changes over time were examined graphically, enabling comparisons with known annual variations such as flooding events. We hypothesized that spring floods would have a greater impact on exotic annuals than summer floods. Consequently, as well as overall tests to examine differences between flooding treatments (by examining the interaction with time as explained above), we also used planned contrasts to compare three groups of treatments: no floods; summer floods; and spring and summer + spring floods. Only significant results from these tests are reported.
Initial analyses gave the anomalous result of greater cover of native perennials in grazed than ungrazed plots in unflooded areas. Examination of pre-treatment data showed that the rhizomic perennial Eleocharis pusilla was patchily distributed and more abundant in grazed than ungrazed plots which were not to be flooded again, and vice versa in plots which were to be flooded in spring. Over all years, cover of E. pusilla was significantly higher in ungrazed plots than grazed plots in the spring and summer + spring flooded plots, while the opposite was observed in the unflooded plots (P = 0·0429). Earlier studies have suggested that this species is not preferentially grazed (Cunningham et al. 1981). Consequently, E. pusilla was excluded and data re-analysed.
- Top of page
- Materials and methods
- Supporting Information
Over 5 years of sampling, a total of 98 species were recorded from 216 1-m2 quadrats, including 60 native and 38 exotic species. The average total cover of all plants was 18% across all plots in all years, ranging from an average of 3% to 48% across the nine, 1-m2 quadrats within a given plot in a given year. The two dominant groups of plants recorded were annual exotic species (23% of cover) and perennial native species (73% of cover). Perennial exotic species made up <1% of cover and are not considered further here. Annual native species had very low abundances (overall average 0·8% cover) and showed no effects of any factors (year, grazing, flooding). Hence, only exotic annuals and native perennials were examined in detail.
As expected, unflooded plots were dominated by terrestrial exotic annuals (especially Hypochaeris glabra and Avena, Lolium, Trifolium and Vulpia species) and flooded plots by aquatic native perennials, particularly Eleocharis acuta and Triglochin procera. The native perennial mudflat species, Eleocharis pusilla, was also abundant in all treatments.
Cover of Annual Exotics
Consistent with the terrestrialization model, flooding reduced the cover of exotic annuals, as shown by the significant interaction between flooding and year (Fig. 1, Tables 1 and 2). Averaged across all years, cover of exotic annuals was significantly greater in unflooded plots (mean 15·6%) than in summer-flooded plots (2·5%), and these, in turn, had significantly greater cover than plots flooded in spring and summer + spring (1·0%; Table 1). There were no exotic annuals in unflooded plots in 1992 when the ‘unflooded’ treatment was inundated by a natural spring flood (Fig. 1). When 1992 data are excluded, the average cover of exotic annuals in the unflooded treatment was 21%, while in the summer-flooded treatment the average cover of exotic annuals was 3·4% (Table 1). Contrary to expectations, summer floods greatly reduced cover of annual exotics in the following spring, although not as much as spring floods (Fig. 1). Grazing had no significant effect on cover of annual exotics, and there were no significant interactions between grazing and other factors (Table 2).
Figure 1. Cover of exotic annual and native perennial plants in each flood regime each year (mean ± 1 SE). All plots were flooded in 1990 before grazing and flooding treatments began. An unplanned natural flood inundated all treatments in spring 1992. Su, summer; Sp, spring.
Download figure to PowerPoint
Table 1. Mean cover of each plant functional group in the spring following each flooding treatment, averaged across both grazing treatments and all years (excluding 1990 when all plots were flooded)
|Unflooded||Summer||Summer + spring||Spring|
|Exotic||Annual/biennial||Terrestrial||11·8 (15·8)||1·8 (2·5)||0·7||0·6|
|Mudflat||3·8 (5·0)||0·7 (0·9)||0·5||0·2|
|Total exotic|| ||15·6 (20·8)||2·5 (3·4)||1·2||0·8|
|Native||Annual/biennial||Mudflat||0·9 (0·8)||0·7 (0·5)||0·8||0·8|
|Perennial||Terrestrial||0·3 (0·5)||0·3 (0·3)||0·2||0·3|
|Mudflat||6·4 (5·8)||3·9 (3·6)||9·3||14·4|
|Aquatic||0·1 (0·02)||0·7 (0·2)||11·4||8·6|
|Total native|| ||7·8 (7·0)||5·6 (4·6)||21·7||24·1|
Table 2. PERMANOVA repeated measures analyses of effects of sub-region (R), flooding (F), grazing (G) and year (Y) on cover of annual exotics, native perennials and native perennials excluding Eleocharis pusilla
|Measure||Source||d.f.||MS||Pseudo-F|| P ||Planned contrastsa|
|Annual exotic cover||R||2||19·1||1·76||0·2554|| |
|F||3||95·8||8·82|| 0·0177 || |
|Error||6||10·9|| || || |
|Y||4||275·6||21·52|| 0·0001 || |
|Y × F||12||158·5||12·38|| 0·0001 ||Y × (N vs. Su): P = 0·0004, Y × (N vs. Sp and Su + Sp): P = 0·0001, Y × (Su vs. Sp and Su + Sp): P = 0·0017|
|Error||32||12·8|| || || |
|G × F||3||0·5||0·05||0·9874|| |
|Error||8||12·0|| || || |
|G × Y||4||11·6||0·29||0·8865|| |
|F × G × Y||12||8·7||0·22||0·9956|| |
|Error||32||39·4|| || || |
|Perennial native cover||R||2||26·9||0·93||0·4426|| |
|F||3||199·7||6·86|| 0·0269 || |
|Error||6||29·1|| || || |
|Y||4||402·3||4·57|| 0·0029 || |
|Y × F||12||147·1||1·67||0·1104||Y × (N and Su vs. Sp and Su + Sp): P = 0·0009|
|Error||32||88·0|| || || |
|G × F||3||58·3||7·04|| 0·0115 ||G × (N and Su vs. Sp and Su + Sp): P = 0·0046|
|Error||8||8·3|| || || |
|G × Y||4||102·8||2·70|| 0·0482 || |
|F × G × Y||12||52·3||1·37||0·2239|| |
|Error||32||38·1|| || || |
|Perennial native cover excluding Eleocharis pusilla||R||2||62·2||2·96||0·1241|| |
|F||3||222·0||10·55|| 0·0108 || |
|Error||6||21·0|| || || |
|Y||4||370·5||4·49|| 0·0030 || |
|Y × F||12||140·3||1·70||0·1017||Y × (N and Su vs. Sp and Su + Sp): P = 0·0004|
|Error||32||82·6|| || || |
|G||1||39·1||19·73|| 0·0019 || |
|G × F||3||25·1||12·70|| 0·0024 ||G × (N and Su vs. Sp and Su + Sp): P = 0·0061|
|Error||8||2·0|| || || |
|G × Y||4||44·5||1·65||0·1889|| |
|F × G × Y||12||41·8||1·55||0·1504|| |
|Error||32||26·9|| || || |
Cover of Native Perennials
Although there was no significant overall interaction between flooding and year for native perennials, there was a significant interaction between year and the contrast between no floods and summer floods vs. spring and summer + spring floods, regardless of whether or not E. pusilla was included in the analysis (Fig. 1, Table 2). This interaction indicates that floods immediately prior to sampling (i.e. spring and summer + spring floods) promoted cover of native perennials to varying degrees in different years, whereas native perennials had low cover in unflooded plots and summer-flooded plots, which were flooded many months before sampling (Fig. 1, Table 1). When the data from 1992 (when all plots were flooded in spring) were excluded, native perennial cover was even lower in the unflooded and summer-flooded treatments (Table 1). There was a significant interaction between grazing and year, which disappeared when E. pusilla was excluded from the analysis. The cover of E. pusilla was higher in grazed than ungrazed plots in 1995, while the reverse was true in all other years. The cover of native perennials also showed a significant interaction between grazing and flooding (Table 2). Grazing significantly reduced the cover of native perennials in plots that were flooded shortly before sampling in spring (i.e. summer + spring and spring floods), but had little effect on the low levels of cover in unflooded plots and plots that were inundated many months before sampling (summer floods; Table 2, Fig. 2). The anomalous finding of increased cover in grazed plots compared to ungrazed plots in the unflooded treatment (Fig. 2) was attributable to greater initial cover of E. pusilla in grazed than ungrazed plots in this treatment.
Figure 2. Cover of all native perennial plants (mean + 1 SE) and of terrestrial, mudflat (divided into Eleocharis pusilla and all other mudflat species) and wetland perennials in grazed (shaded) and ungrazed (unshaded) plots in each flood regime averaged across all years. The error bars are for total cover of all native perennial plants.
Download figure to PowerPoint
All three functional groups of native perennials (terrestrial, mudflat and aquatic species) responded to treatments in a similar way (Table S1). The cover of each group varied significantly among years, with significant interactions between year and the contrast between no floods and summer floods vs. spring and summer + spring floods, and between grazing and the contrast between no floods and summer floods vs. spring and summer + spring floods. Aquatic species also showed a significant overall interaction between year and flooding (Table S1). Cover of native perennial mudflat and aquatic species was greatest in plots flooded in summer + spring and spring (Fig. 2, Table 1). These plots were inundated shortly before spring sampling. Grazing reduced the cover of both functional groups in plots flooded in summer + spring and reduced the cover of mudflat plants in plots flooded in spring. Grazing did not affect cover in unflooded and summer-flooded plots, where cover of both functional groups was low (Fig. 2). Notwithstanding significant statistics (Table S1), the cover of terrestrial native perennials was low (<0·4%) in all treatments in all years, and it was difficult to interpret responses ecologically.