between-site differences in e. plantagineum DeMOGRAPHIC PARAMETERS
The analyses in this study broke down the life cycle of E. plantagineum into specific demographic parameters and estimated these both in the native and invaded ranges, thereby allowing for a comparison of demographic processes between the sites. The largest difference between the sites was the up to five times greater seedling establishment fraction at Canberra than at Evora. This result indicates that the establishment of E. plantagineum seedlings from the seed bank in Mediterranean Europe is considerably more limited by factors that are either not present, or weaker in effect, in Australia. The availability of germinable seeds and safe sites for establishment are the two factors that may limit recruitment in plant populations (Eriksson & Ehrlén 1992). A comparison of the proportion of seeds germinating and the establishment of seedlings between Canberra and Evora can be used to distinguish between these two possibilities. The proportion of the seed bank lost during the germination period was broadly similar between Canberra and Evora (Fig. 5), but the much lower seedling establishment fractions at Evora show that a smaller proportion of these presumably germinating seeds survived to produce established seedlings. This suggests that the frequency of safe sites for establishment (Harper 1977) is lower at Evora than at Canberra.
Numerous mechanisms may result in a lower number of safe sites for E. plantagineum establishment at Evora than at Canberra. It is possible that the species-rich annual community present in Mediterranean annual pastures, often well over 100 species (Fernandez-Alés, Laffarga & Leiva 1991; Fernandez-Alés, Laffarga & Ortega 1993), presents significant competition for E. plantagineum for establishment space during autumn, leading to a high mortality of germinating seeds at Evora. The situation in Australia, where improved pastures tend to be species poor (Kemp & Michalk 1994), may lead to lower intensities of competition for establishment space, allowing E. plantagineum seedlings to dominate the available bare ground (Smyth et al. 1992). Other possible mechanisms for the lower seedling establishment fractions at Evora concern the presence of seedling predators and pathogens or allelopathic effects limiting the establishment of germinating seeds more so than at Canberra (Burdon & Shattock 1980; Brown 1994; Hanley, Fenner & Edwards 1995; Clear Hill & Silvertown 1997).
In contrast, the lower seedling survival rates measured at Canberra than at Evora indicate that not all stages of E. plantagineum's life cycle are favoured in the Australian environment. Similarly high mortality amongst juveniles in Australia of between 41–97% (Burdon, Marshall & Brown 1983) and 20–65% (Smyth, Sheppard & Swirepik 1997) have been recorded previously. In the Australian environment, dry periods subsequent to germination-inducing rains, particularly during summer, and the presence of livestock grazing during summer and autumn can result in significant juvenile mortality (Burdon, Marshall & Brown 1983). At Evora, a greater reliability of follow-up rains during autumn (Grigulis 1999) allows those seedlings that have established a higher probability of surviving to flower.
Plant fecundity was in general very similar between the sites. This result is in contrast to the observation that, in alien environments, plants tend to be more vigorous and taller, producing more seeds than in the native distribution (Blossey & Notzold 1995). This latter phenomenon has been attributed to releases from natural enemies and the subsequent increases in plant performance, and possibly the redeployment of resources used for herbivore defence to growth in more favourable environments (Blossey & Notzold 1995).
The 60–70% seed bank survival rate was similar between the sites and is high compared to other published estimates for weed species. The average annual rate of seed decline for weed species has been estimated to be approximately 50% (Snaydon 1980). The presence of such a strongly persistent seed bank suggests that the successful control of E. plantagineum in Australia will require long-term control measures as significant quantities of seed will remain in the soil for considerable periods of time.
Although highly variable, the seed bank incorporation rate was three times greater at Canberra than at Evora. Field observations indicate that the harvesting of E. plantagineum seeds by ants (Messor spp.) at Evora may be an important factor in the greater loss of seeds there. Seed harvesting of this magnitude was not observed at Canberra (Grigulis 1999).
This study suggests that the greater success of E. plantagineum populations in Australia than in Mediterranean Europe is due to higher rates of seedling establishment from the seed bank and a higher rate of incorporation of fresh seeds into the seed bank. Mechanisms that determine the fate of plants often act during the period of the life cycle encompassing seed dispersal, germination and establishment (Grubb 1977; Harper 1977). Noble (1989) also identified the phases of reproductive losses between flowering and the seed pool and the rate of seedling establishment as those which may be expected to change for a species entering a new environment. The results for E. plantagineum correspond well with these conclusions and provide further evidence of the often critical importance of the recruitment process in the determination of population size and in determining the success of a plant invasion. More recently, studies on annual weed species have suggested that some population demographic parameters may remain remarkably constant both spatially and temporally (Freckleton & Watkinson 1998; Lintell Smith et al. 1999). In these two studies those parameters tending to remain constant were those relating to yield–density relationships, such as seed production, where variations in individual plant performance were compensated for at the population level through density-dependent effects. In contrast, changes in the rates of emergence and mortality of seeds in these studies, such as those created by cultivation or environmental variation, were critical in determining changes in population size. These results, together with the results for E. plantagineum, suggest that the buffering of population size through density-dependent recruitment is less able to compensate for losses than the buffering of biomass and seed production through yield–density relationships, and thus, for annual weed species, variability in population size may be generally driven by variations in the rates of seed germination and emergence.
Effect of the grazing and pasture competition treatments on the demographic parameters
Grazing considerably increased the seedling establishment fraction of E. plantagineum both at Canberra and at Evora. Increases in seedling establishment with grazing and other disturbances have been reported frequently (Goldberg & Werner 1983; Rapp & Rabinowitz 1985; Osterheld & Sala 1990; Milton 1995) and have been attributed to decreases in adult : seedling competition (Fenner 1978) and the removal by disturbance of inhibitory litter layers (Gross 1980). In the absence of grazers, Mediterranean annual communities are commonly dominated by a few species of tall, competitive, large-seeded annual grasses the litter of which often inhibits the regeneration of subordinate species (Noy-Meir & Briske 1996). Lower E. plantagineum seedling establishment in ungrazed plots at both sites was probably due to the inhibitory effects of such thick litter layers that developed in the absence of grazing (Grigulis 1999). Under the presence of grazing the seedling establishment fraction was considerably greater at Canberra than at Evora. This suggests that pastures responded differently to grazing between the sites, with grazed pastures at Canberra providing more safe sites for establishment than grazed pastures at Evora.
Irrespective of the grazing treatment, the survival of E. plantagineum seedlings from establishment until flowering was always lower at Canberra than at Evora, indicating that grazing was not a factor limiting seedling survival at Evora more than at Canberra. Similarly, grazing did not consistently decrease fecundity at Evora more so than at Canberra. However, grazing intensities can vary greatly from season to season and from site to site, with reported values for seed losses due to grazing for E. plantagineum in Australia ranging from 45% to 98% (Piggin & Sheppard 1995; Smyth, Sheppard & Swirepik 1997). Consequently, results from this experiment of a two-site comparison over two seasons must be treated with caution.
Diffuse pasture competition reduced the fecundity of individuals of E. plantagineum in both grazing treatments equally (by 30–45%) at both Evora and at Canberra. Consequently, there is also no evidence that pastures in Mediterranean Europe impose greater competitive effects on established plants than pastures in Australia, and hence reduce E. plantagineum performance in Europe. However, interactions between natural enemies and pasture competition can be important in determining E. plantagineum fecundity (Sheppard, Smyth & Swirepik 2000) and these interactions may take different forms between a plant's native and invaded ranges (Sheppard 1996).
The seedling establishment fraction and the seed bank incorporation rate were the most important parameters in determining the greater population abundances of E. plantagineum at Canberra compared with Evora. This suggests that reducing these two parameters may produce the greatest decreases in E. plantagineum abundances. Additionally, if it is possible to determine what limits both of these parameters more in Mediterranean Europe than in Australia, this may indicate ways by which these parameters could be lowered in Australia. The low seedling establishment rates of E. plantagineum observed in ungrazed plots, where the plots were dominated by a thick cover of vegetation and litter, suggests that reducing bare ground in autumn to decrease the space available for E. plantagineum establishment may reduce rates of E. plantagineum seedling establishment. Indeed, E. plantagineum frequency is generally lower in the predominantly perennial pastures of Tasmania than in the more annual pastures of south-east and south-west Australia (Friend 1991). Similarly, Kemp, Dowling & Millar (1990) found that a 3-month rest from grazing in autumn or winter decreased the proportion of weedy annual grasses in Phalaris-based pastures. Forcella & Wood (1986) also showed that increases in the basal cover of a pasture decreased the density of annual thistle species, illustrating the efficacy of maintaining pasture cover as a control method for annual weeds. While it is unlikely that farmers can leave substantial areas ungrazed to allow such control, the maintenance of significant vegetation cover during the critical recruitment period may be achieved by strategically timing decreased stocking rates to increase pasture cover before autumn and into winter and preventing over-grazing during this period, or through the adoption of rotational rather than continuous grazing regimes. Such a grazing strategy, coupled with the several biological control agents now established in Australia that are reducing E. plantagineum survival and seed production (Piggin & Sheppard 1995), and the wider adoption of spray-grazing, a sublethal herbicide application followed by a short period of intense grazing in autumn to reduce the survival of E. plantagineum rosettes (Piggin 1979; Smyth & Sheppard 1996), provides an integrated approach for the management of E. plantagineum in Australian pastures.
Assessment of the demographic approach
The fact that introduced species often vary in their behaviour in different regions provides an on-going experiment in biological invasions. Such comparative studies offer opportunities to advance our understanding of the invasion process in general, as well as a means to manage ecosystems better to resist such invasions (Kruger et al. 1989). Comparing the population behaviour between two such environments using demographic methods allows the localization of the mechanisms important in enhancing the abundance of the species in one environment, as opposed to the other, by the isolation of differences between the populations in particular life-history stages. The approach used in this study involved the coupling of comparative and demographic approaches and proved to be useful in identifying the demographic mechanisms increasing the abundance of E. plantagineum in the pasture community at Canberra compared with at Evora. It is now possible to test hypotheses derived from this approach using focused manipulative experiments. Emphasis in future research on plant invasions should move away from purely correlative studies, such as the invasibility of an ecosystem and attributes such as its diversity, or the invasiveness of a species and lists of traits it may possess (Lavorel, Prieur-Richard & Grigulis 1999b). A better understanding of such relationships will be gained from an emphasis on the underlying mechanisms of biological invasions (Lavorel, Prieur-Richard & Grigulis 1999b) using approaches such as those used in this paper (Sheppard 1999).
However, estimating the demographic components of a species’ life history can be difficult as survival, growth and reproduction in plant populations can vary considerably from site to site and from year to year (Damman & Cain 1998). Echium plantagineum populations were followed for only two seasons at each site, thereby raising the question of what proportion of the possible variability in population demographic parameters has been captured at each site. Similarly, the generality of the mechanisms limiting E. plantagineum abundance at Evora compared with at Canberra is unknown, as only one E. plantagineum population was studied in both the native and invaded ranges. The use of only one site in each of the regions may bias the results if the chosen population in each region does not behave similarly to other populations in that region. Ideally, a number of populations of both high and low abundance of E. plantagineum should be studied both in the native and invaded ranges to determine the generality of differences between the two regions. Despite these shortcomings, a comparative demographic approach has the potential to generate hypotheses as to why a species is invasive in a particular environment, the generality of which can then be tested over a much wider range of environments.