Seedling recruitment patterns over 4 years in an Australian perennial grassland community with different fire histories

Authors


J. Morgan (tel. + 61 39479 2226; fax + 61 39479 1188; e-mail j.morgan@latrobe.edu.au).

Summary

  • 1Seedling recruitment was followed from 1993 to 1996 in four remnant Themeda triandra grassland sites in south-eastern Australia subjected to different long-term fire histories (1-, 2-, 4-year fire intervals, unburnt > 10 years). Multivariate analyses were used to separate the effects of site and year on recruitment and community dynamics.
  • 2Exotic species (mostly annual monocots such as Aira spp., Briza minor and Juncus capitatus) dominated the seedling flora at all sites in all years. Seedlings of most native species were absent or rare in all years, despite their abundance in the standing flora.
  • 3Seedling recruitment (richness and density) in the long unburnt grassland was significantly less than in burnt sites in all years. Hence, the seedling floristic composition of this site was substantially different from all other sites on PCA axis 1, which did not differ in all years.
  • 4Amongst burnt sites, some differences in the annual seedling cohort were evident on PCA axes 2 and 3, but these were due to small differences in the relative abundance rather than the composition of recruiting species. Seedling recruitment was not cued or promoted by fire but rather, seedling density was influenced by, and interacted with, yearly (presumably rainfall) variation.
  • 5Amongst the native species seen as seedlings, yearly emergence occurred in some cases (14 of 24 species) but always at low densities (< 10 seedlings/0.25 m2). Seedling survival amongst native species varied from nil (0%) to high (> 60%), with most mortality occurring in early summer.
  • 6The temporal and spatial variation observed in seedling regeneration suggests that long-term native species coexistence may be promoted by differences in the ‘regeneration niche’. Seedling regeneration, however, has minimal impact on the short-term dynamic of this community, whose conservation requires maintenance of the existing ‘bud- and tuber-bank’ of native species.

Introduction

The reasons why some plant communities have high species diversity have long interested ecologists and have given rise to many hypotheses about the mechanisms of species coexistence (e.g. Grubb 1977; Grime 1979; Warner & Chesson 1985; van der Maarel & Sykes 1993). Grubb's (1977)‘regeneration niche’ theory, whereby different regeneration patterns permit coexistence of a wide range of species even if they have similar adult physiological requirements and tolerances, has found particular favour in studies of grasslands (e.g. Goldberg & Werner 1983; Hobbs & Mooney 1985; Rusch & van der Maarel 1992; Rusch 1992; Williams 1992; Kotorova & Leps 1999).

Temperate, species-rich grasslands have only recently been recognized in south-eastern Australia and are highly endangered (McDougall et al. 1994), such as the volcanic plains grasslands of western Victoria, which have been reduced to less than 1% of their former extent due to agricultural development and urbanization. Annuals comprise less than 10% of the original native flora (Willis 1964) but many sparse, native perennial hemicryptophytes and geophytes occur in the interstitial spaces between tussocks of the dominant C4 grass, Themeda triandra. Species richness can be high, even at small scales (12 species/0.01 m2, Morgan 1998a, vs. 58 species/100 m2, McDougall et al. 1994). Nowadays, exotic annual species also contribute substantial numbers of species to these grassland remnants (Morgan 1998a), perhaps due the proximity of agricultural land, and because they occupy a niche not well utilized by native species (Johnstone 1986).

Remnants of this community along roadsides and railway enclosures are particularly species rich, presumably because they have not been heavily grazed by domestic stock nor improved by superphosphate addition (Stuwe & Parsons 1977), but have been burnt in summer, mostly at 1–3-year intervals, to prevent the spread of wildfire. This has prevented the competitive exclusion of many interstitial species by the dominant grass (Stuwe & Parsons 1977; McDougall 1989) and is thought to mimic more closely the aboriginal burning regimes to which these grasslands would have been exposed over thousands of years (Stuwe & Parsons 1977). Such remnants grasslands are therefore considered to be the best examples, or ‘working models’ (sensuScarlett 1994), of the original indigenous plant community and are accorded the highest conservation priority (McDougall et al. 1994).

There are few data, however, about the spatial and temporal dynamics of any of the interstitial species in this community and hence, little information to guide conservation management. Frequent burning (i.e. 1–3-year intervals) may be needed for recruitment, and hence species coexistence, in a community where > 70% of species are non-clonal and therefore dependent on seedlings for reproduction (McIntyre et al. 1995; Morgan 1998a). In the absence of burning, accumulation of litter and deep shading by the canopy (Lunt 1995; Morgan 1997; Morgan & Lunt 1999) may inhibit recruitment of some species while others disappear when a thick, continuous canopy develops (e.g. Scarlett & Parsons 1990) and cannot then recover after a subsequent fire because of the paucity of long-lived seed in the soil (Lunt 1990; Morgan 1995a,b; 1998a).

Information on regeneration in this community is limited to species-specific studies of endangered taxa (e.g. Gilfedder & Kirkpatrick 1994; Morgan 1995b) and the dominant grass (Themeda triandra (McDougall 1989)). Only Lunt (1990) has documented community regeneration, albeit in a formerly grazed and long-unburnt grassland, where he found minimal regeneration of native species following an autumn fire, but abundant regeneration of exotic annual monocots that were derived from a soil seed bank built up over an 80-year period of stock grazing.

Hence, the aims of this study were to (i) document the amount of, and temporal variation in, seedling recruitment in species-rich grasslands that are subject to different management regimes, and (ii) follow seedling survival for major cohorts of native species to determine whether successful seedling establishment occurs under the existing management regimes.

No attempt was made to identify specific microsite requirements for germination as has been done for limestone grasslands by Rusch & Fernandez-Palacios (1995). Data on the spatial heterogeneity and effects of canopy gap size on recruitment are presented elsewhere (Morgan 1998b).

Methods

Study area

The study was undertaken on the flat to undulating volcanic plains of western Victoria, Australia, where most of the lava flows date from the Upper Cainozoic. Soils are generally acidic to neutral (pH 6–7), grey to red-brown, clay to clay loams with an A2 horizon of buckshot ironstone (McDougall et al. 1994).

The climate is temperate, with cool winters (mean maximum temperature of the coldest month (July) is 4 °C) and warm summers (mean maximum temperature of the warmest month (January) is 26 °C) (Bureau of Meteorology & Walsh 1993). Annual rainfall is approximately 610–710 mm at all sites, with a minor peak in spring (McDougall et al. 1994). Evaporation exceeds precipitation from October to April, hence producing a summer water deficit in most years (Land Conservation Council 1976). A distinct ‘autumn break’ also occurs in most years. This is characterized by substantial heavy rainfall over a short duration in late autumn that rapidly saturates the upper soil profile, which had previously been dry due to ineffective autumn showers (Morgan 1995b).

Monthly rainfall during the study varied substantially from one year to the next, but similar patterns were observed at all sites (Fig. 1). During 1994 below-average rainfall was recorded at all sites in all months except February (i.e. 77–84% of average annual rainfall fell for the year) whilst otherwise rainfall was generally close to or above average in most months (Fig. 1), particularly in the period June–September when germination is most pronounced.

Figure 1.

Monthly rainfall for the period January 1993 to September 1996, as a percentage of the long-term average, recorded from the nearest meteorological station to each of the three study sites.

Each study site was chosen on the basis that it was dominated by Themeda triandra and that the recent burning history was well documented (McDougall et al. 1994, and personal communications to local landholders) and had been implemented on a consistent basis for > 10 years. Four sites were chosen, with fire cycles of 1-, 2- or 4-years, or unburnt for > 10 years (Table 1). These regimes were maintained during the study period. Hence, each site differed in that it was exposed to a different number of fires during this study (either 4, 2, 1 or 0). No site had a history of recent (i.e. within 20 years) stock grazing or mowing.

Table 1.  Study site characteristics and fire history during the period of observation 1993–96. Standing biomass was determined in 1994 by harvesting 12, 0.25 m2 quadrats to ground level and drying at 80 °C for 48 h
Site name/codeLocationLong-term fire historyAnnual rainfall (mm)Site tenureMaximum standing biomass (g mminus;2) Total site richnessFire history during study
Derrinallum (D-1)143°7′S, 37°55′E1 year intervals for > 45 years630Road verge25571 native, 22 exoticBurnt 1993 1994, 1995, 1996
Karrabeal (K-2)142°10′S, 37°36′EMostly 2 year intervals for > 40 years. Last burnt summer 1992710Road verge27567 native, 19 exoticBurnt 1993, 1995
Bannockburn (B-4)144°9′S, 38°2′E3–4 year intervals for 15–20 years. Last burnt autumn 1991610Railway enclosure31546 native, 11 exoticBurnt 1995
Karrabeal (K-U)142°12′S, 37°37′EUnburnt > 10–15 years, probably much longer710Road verge67547 native, 16 exoticUnburnt

Seedling emergence

At each site in March 1993, 10 0.25 m2 quadrats were placed at 1 m intervals along each of three randomly located 10 m transects in visually homogenous vegetation and permanently marked with metal spikes in opposing corners. Transects were located within 20–200 m of each other. Whilst replication of transects and quadrats occurred within each of the four sites, there were no replicate sites for the different burning regimes under study and the design of this study is thus pseudoreplicated (Hurlbert 1984). General conclusions can therefore be drawn about patterns of variation among sites, but not about the effects of burning regimes per se.

At approximately 6 week intervals in winter–spring (June–October) during 1993–96, all seedlings within each quadrat were identified, where possible, and counted. A quadrat frame divided into 10 cm × 10 cm subquadrats was used to count seedlings accurately. Some seedlings, particularly those of small dicots, could not be identified to generic level before dying. Many other seedlings could not be identified to species level and were therefore grouped (e.g. Aira spp., Isolepis spp.).

Twelve quadrats per site were selected at random and each year in December from 1994 to 1996 all perennial species rooted within them were recorded to enable comparisons between the standing flora and the seedling flora.

Seedling survival

The survival of native perennial dicot seedlings in each of the major seedling cohorts from 1993 to 1995 was followed by permanently marking their locations with numbered bamboo stakes. Themeda triandra, the major perennial monocot seedling, was not followed because of difficulties with accurate long-term identification of seedlings and small tillers. A major seedling cohort was defined as one where the mean seedling density at a particular site (n = 30 quadrats) exceeded one seedling per 0.25 m2. At approximately monthly intervals after germination, for at least 7 months (and hence, encompassing a dry summer period), the survival of each individual was noted and the number of green leaves was recorded as a measure of growth. Some seedlings that became leafless during summer were scored as dead before subsequently resprouting in the following autumn. Hence, seedling survival is more correctly viewed as ‘apparent’ survival.

Data analysis

The number of species (analysed using the categories native and exotic) and the total number of seedlings (native, exotic, native annual, native perennial, exotic annual, exotic perennial) per quadrat (n = 30) were compared between and within sites using multivariate repeated measures analysis of variance (manova; Winer et al. 1991). This approach has been used to account for the multivariate nature of the data (site × year) using data collected from repeated measurements on the same experimental plots (Kent & Coker 1992; Guo & Brown 1996; Lesica & Steele 1996). Richness data were square root (x + 0.5) and density data log (x + 1) transformed prior to analysis. All 30 quadrats per site were used in each analysis. The Bonferroni multiple comparison procedure was used to guard against type I errors (Kirk 1968).

Jaccard's coefficient of similarity (Sj) (Kent & Coker 1992) was calculated to quantify the overlap in perennial species composition between the seedling and standing flora from 1994 to 1996 at each site in each of 12 permanent quadrats. Sj values were arcsine transformed prior to manova with Bonferroni multiple comparisons.

Principal components analysis (PCA), using the program canoco (ter Braak 1992), was used to characterize the floristic composition of the seedling flora across all years and sites. PCA is a robust way to ordinate all samples if, as here, the common species are shared amongst most of the samples in the analysis but they vary substantially in their relative abundance (Guo & Brown 1996). Data used were log(x + 1) transformed density counts of species summed over the 10 replicate quadrats found in each of the three transects at each site in each year. This gave a total of 48 (12 transects × 4 years) censuses that were subjected to PCA. Only species with a mean density of ≥ 0.1 seedlings per quadrat were included in the analysis.

The first three PCA axis scores were subjected to manova with Bonferroni multiple comparisons to determine the effects of year and site on seedling composition (Guo & Brown 1996). The effects of climatic variation on the seedling flora are indicated by coordinated chronological changes in PCA scores across all sites, whilst the effects of site are indicated by the separation of each site, either over time or in direct response to each burn (Guo & Brown 1996).

Results

General patterns of recruitment across all grasslands

The seedling flora was dominated by exotic species (both in richness and density) in all years and at all sites. Exotics outnumbered natives by 2 : 1 species and by 10 : 1 seedlings in most years at all sites (Fig. 2). At burnt sites, exotic annual monocots (e.g. Aira spp., Briza minor, Juncus capitatus) dominated the seedling flora in all years, whereas at the unburnt site, exotic annual monocots (e.g. Critesion murinum) and exotic perennial dicots (e.g. Hypochoeris radicata) were the most common species as seedlings (Table 2). The similarity between the perennial species in the seedling and standing floras was extremely low in all years at all sites (Sj < 15 for burnt sites and Sj < 25 for the unburnt site; Fig. 3), as most perennial native species were entirely absent from the seedling flora in all years. Similarity did not differ between burnt sites (P > 0.05) in any of the 4 years of the study.

Figure 2.

Mean (±1 SE) (a) richness and (b) total number of seedlings per 0.25 m2 of native (▪) and exotic (□) species from 1993 to 1996 at four Themeda triandra grasslands subjected to different long-term fire histories (1-, 2-, 4-years, unburnt). The annually burnt site was burnt in summer (January) in all years; the 2-year site was burnt in summer 1993 and 1995; the 4-year site was burnt in summer 1995.

Table 2.  Mean number of seedlings per 0.25 m2 (n = 30) of the most common (> 0.50 seedlings/0.25 m2 in any year) annual and perennial exotic and native species to germinate from 1993 to 1996 at four Themeda triandra grassland sites in south-eastern Australia (see Table 1 for an explanation of the site codes and burning histories). Life forms were assigned using McIntyre et al. (1995) as a guide. T = therophyte, G = geophyte, H = hemicryptophyte, C = chamaephyte
SpeciesLife formSite/Year
D-1K-2B-4K-U
93949596939495969394959693949596
Exotic Annuals
*Aira spp.T85.3069.0318.53223.93119.6377.43109.4754.17247.5344.23205.23245.53 1.470.83 3.53 0.33
*Briza maximaT  0.83   0.33  0.07     2.03 0.20  1.00  1.90 0.03   
*Briza minorT20.5716.8020.03 10.53 12.9319.23 18.63 2.70 51.3311.80 48.27 24.8712.100.30 1.13 0.30
*Cicendia quadrangularisT 6.50 0.60 0.60  3.80 40.1030.03  5.03 1.10  5.43 0.23  0.07  0.10 0.330.07  
*Critesion murinumT             2.202.0015.8713.43
*Juncus capitatusT 5.00 1.17 9.13 27.90  1.60 0.80 23.13 0.87  2.27 0.10 52.20  5.30   0.60 
*Moenchia erectaT 6.53 2.27 0.20  1.33    0.30         
*Parentucellia latifoliaT   0.03  0.17  5.27 3.27 11.63 0.03        
*Sonchus oleraceusT 3.47 1.50 0.33  0.47  0.07   0.07  0.90 1.10  0.50  0.30   0.10 
*Trifolium dubiumT  0.33    1.57 1.03  3.10 0.17     0.03  
*Trifolium spp.T 1.00 0.03 0.63  0.10  0.10 4.97  1.13 0.30    0.30  0.100.03 0.47 
*Trifolium subterraneumT 1.23 0.30 0.03  0.07  4.80 0.10 14.93         
*Vulpia bromoidesT26.9026.3319.47 27.80 10.93 9.17  3.30 9.77 10.50 0.63  0.33  0.50 2.070.37 0.30 0.03
Exotic Perennials
*Holcus lanatusH  0.03 0.03  0.07 0.13 0.10 0.03  0.13 0.03 0.13 2.500.97 7.07 1.17
*Hypochoeris radicataH 0.63 0.50 0.67  0.47 1.73 1.00 0.47 0.07 0.77 0.30 0.43 0.10 2.174.9315.30 1.43
*Romulea roseaG 1.10 0.60   0.03 0.33  3.13 0.57 3.33 0.03  6.70    0.33
Native Annuals
Crassula sieberianaT 1.50 4.43 2.47 10.67 0.03 0.13 0.17  0.03 0.53 4.83 0.77    
Isolepis spp.T14.57 3.43 1.37  4.87 22.20 8.93 32.90 4.27 4.90 0.23 21.97 5.97 1.13   
Juncus bufoniusT 1.03  1.47  2.10 0.50 0.03 0.03 0.17 0.03 0.03 0.03  0.03   
Juncus spp.T  1.63  20.70 33.50  0.80  0.57      
Sebaea ovataT 3.40 0.33 1.43  1.17 3.63 2.43 1.53 1.07 4.50  0.87     
Native Perennials
Eryngium ovinumH 2.90 0.23 0.03  0.03    0.13   0.03    0.03
Leptorhynchos squamatusH 5.70  0.10  0.50 0.03 0.03  3.63 0.97 0.03 1.73    
Leptorhynchos tenuifoliusH 2.03  0.73  0.97         0.03   
Pimelea curvifloraC         0.47  1.60 0.07    
Plantago gaudichaudiiH   0.10       0.17 3.47 0.07    
Schoenus apogonH 0.47 0.03    0.07   2.20 0.03      
Themeda triandraH 1.23 6.23  5.73 0.87 3.43  0.67 0.20 0.33  2.50 0.23  0.03
Figure 3.

Mean (±1 SE) index of similarity (Sj) between the perennial species in the standing flora and the seedling flora from 1994 to 1996 at four Themeda triandra grassland sites subjected to different fire histories.

Differences in seedling numbers and composition across sites

The unburnt grassland had significantly fewer species, both native and exotic (P < 0.001), and lower densities of native annuals, native perennials and exotic annuals (P < 0.001) in the seedling flora than all burnt sites in all years (Fig. 4). As a result, the floristic composition of the seedlings at the unburnt site differed significantly from all other sites on PCA axis 1 (P < 0.001; Fig. 5) and differences on PCA axes 2 and 3 were also evident throughout the study (Fig. 5) for this site, where there was little change in seedling composition.

Figure 4.

Mean (±1 SE) total number of seedlings of (a) annuals and (b) perennials of native (▪) and exotic (□) species from 1993 to 1996 at four Themeda triandra grasslands subjected to different long-term fire histories (1-, 2-, 4-years, unburnt).

Figure 5.

The floristic composition of the seedling flora of four Themeda triandra grasslands with different long-term burning histories (1-year: ◆; 2-years: □; 4-years: ▪; unburnt: ▵) on PCA axes 1, 2 and 3 from 1993 to 1996. These axes account for 35%, 11% and 9% of the total floristic variation observed. Co-ordinated axis scores from one year to the next indicate environmental (presumably climatic) effects, whilst separation of PCA scores in canonical space suggest site effects.

There was no significant difference (P > 0.05) between the burnt sites in the number of exotic annual seedlings emerging in any year (Fig. 4). There was, however, much within-site and between-year variability. Emergence of other species groups was generally too sparse in each year to distinguish differences between sites (Fig. 4). Native perennial species were uncommon as seedlings in all years at all burnt sites and were significantly fewer in all years at the biennially burnt site (Fig. 4). The only native species seen as seedlings were already present in the vegetation. Hence, fire-cued germination was mostly not detected in those grasslands where the burning frequency occurred at > 1 year intervals, nor did annual fire consistently promote greater germination over sites burnt less frequently.

Floristically, the seedling composition of all burnt sites did not differ from one another over the 4 years of the study on PCA axis 1 (P > 0.05; Fig. 5). On axis 2, however, the 4-year burn site was significantly different (P < 0.001) from other sites in both the years before (1993, 1994) and after (1995, 1996) burning. On PCA axis 3, the annually burnt site differed significantly during two of the 4 years from all other sites (P < 0.001), which did not differ from one another (P > 0.05; Fig. 5). Floristic differences were due almost entirely to changes in the abundances of infrequent species, and some common species, rather than due to large differences in the suite of species emerging at each site in response to individual fires.

Temporal changes in the seedling flora

Seedling numbers varied significantly from year to year within and between all burnt sites (P < 0.001; Fig. 4), primarily due to large fluctuations in the densities of exotic annuals present. Other groups of species also varied from year to year, but not with the same magnitude. Because few seedlings in any category emerged in any year at the unburnt site, this grassland appeared to be least affected by annual fluctuations in the environment.

manova on principal components scores of the first three PCA axes (which account for 55% of the total variation), indicated highly significant effects of year on the floristic composition of the seedling flora of each site (Fig. 5; P < 0.001). Co-ordinated chronological changes in PCA scores on axes 1, 2 and 3 from one year to the next, indicative of the effects of an over-riding background environmental variation, were evident for all sites from 1993 to 1994 during a year of below-average rainfall (Fig. 5). Other co-ordinated changes were evident and showed consistent effects regardless of the site treatment (Fig. 5). Significant year × site interactions were found, however, in manovas performed on both seedling density data (P < 0.001 for all categories) and PCA scores (P < 0.001 for all axes), suggesting that the effects of site treatments in any one year depend on climate (Fig. 5). Thus, PCA axes scores may change in a related direction and with similar magnitude from year to year (in response to climatic variation) but the sites often remained separated in canonical space (as a result of differential effects of site history, Fig. 5).

Native perennial seedling recruitment patterns

The total number of native perennial seedlings and the number of such species germinating were low for all sites in all years. The unburnt site had almost no native seedling recruitment. Only 12 perennial native species regenerated by seed at the annually burnt site over the 4 years of this study, six at the biennially burnt site, whereas 20 species germinated in total at the 4-year burn site, many at exceedingly low levels (Table 2). In any one year, only 1–12 perennial native species were observed in the seedling flora of each site.

Amongst burnt sites, the same species tended to regenerate by seed across all sites (e.g. Leptorhynchos squamatus). Regeneration occurred in small quantities for all species observed in the seedling flora and, for many species, often occurred in most years, regardless of the stage in post-fire development (Table 2). For each year, one or two species ‘dominated’ the seedling cohort at a particular site or sites (Table 2). For example, Leptorhynchos squamatus dominated both the 1-year and 4-year burn sites in 1993, whereas Themeda triandra dominated the seedling cohort of all sites in 1996 (Table 2). A ‘dominant’ species in the seedling flora was typically not restricted to germination only in that year but rather emerged in most years and could again become the dominant at a later time in the same grassland (e.g. L. squamatus). No species was restricted to germination in the year immediately following a fire, although the densities of some species appeared to be greatest at this time (e.g. Pimelea curviflora at the 4-year burn site during 1995, immediately post-fire; Table 2).

Native dicot species seedling survival

Seedling survival was variable, but species could be separated into two main groups. Most of the first group showed frequent recruitment and had seedlings that ‘dominated’ the seedling population regularly (e.g. Eryngium ovinum, Leptorhynchos squamatus, L. tenuifolius;Table 2). Seedlings of these species were short-lived in all years (i.e. they exhibited 85–100% mortality) at all sites and successful recruitment was therefore generally minimal (Fig. 6). Mortality, which was most pronounced at the onset of summer drought (January–March), did not appear to be due to fire at any site in any year. Some of these species, however, survived over summer as dormant, leafless seedlings (for over 3 months) before regrowing in autumn (e.g. L. squamatus, Fig. 6). The second group (e.g. Pimelea curviflora, Plantago gaudichaudii) consisted of species that recruit less often as ‘dominants’ but have much greater potential for longer-term seedling survival (e.g. > 25%; Fig. 6), showing slow initial growth in the first year and an ability to survive for extended periods over summer as very small or leafless plants (Fig. 6).

Figure 6.

Percent seedling survival (□) and growth of surviving seedlings (mean number of leaves per surviving seedling: ▪) for major seedling cohorts of native perennial dicots. The species, site, observation period, total number of seedlings in the cohort and timing of fire (↓) are indicated for each of the seven observed cohorts. An increase in survival indicates seedlings resprouted after summer (March) after having been completely leafless for some months.

Discussion

Over the 4 years of this study, some general patterns of seedling recruitment are apparent for species-rich Themeda triandra grassland remnants in western Victoria. Most significantly, the seedling flora (in terms of richness and density) of all sites in all years was dominated by exotic annual monocots, despite these remnants being the richest surviving examples of the original vegetation type (McDougall et al. 1994; Scarlett 1994). Exotic annual species dominate the soil seed bank at all species-rich sites (Morgan 1998a) and hence, their predominance in the seedling flora is not entirely unexpected. They are ‘non-specialist’ germinators (Morgan 1998c) that presumably take advantage of the predictable seasonal rainfall and frequent spatial and temporal vegetation disturbances that characterize this vegetation, i.e. there is a ‘stable invasion window’ (sensuJohnstone 1986). Given the relative paucity of native annuals in the natural community (Willis 1964), exotic annuals may have exploited an available ‘niche’ that is not well utilized by native species.

In contrast, native seedlings were observed in only small quantities and only represented a restricted number of the species present at a site. Many species were never seen as seedlings and a further 21% of the perennial species recorded only one seedling in a single year. Whilst some of these species can spread vegetatively (e.g. Asperula scoparia), many are (presumably) non-clonal and therefore depend on seedling recruitment to maintain populations in the long term (e.g. Goodenia pinnatifida). The ‘storage effect’ model of Warner & Chesson (1985) may need to be invoked as a mechanism to help explain persistence of the many species with infrequent recruitment. Under this model, infrequent but highly successful recruitment can be ‘stored’ into the adult population and allow long-term persistence of the species at the site during subsequent periods of low recruitment. Substantial recruitment will necessarily be restricted to a year following average or high seed production to compensate for the absence of a persistent seed bank for most species (Lunt 1994; Morgan 1998a). This usually corresponds to a short period (i.e. 6–12 months) post-fire when flowering is promoted in many species (Lunt 1994). This is in stark contrast to the general seedling recruitment patterns observed in many heathland and woodland communities elsewhere in Australia (e.g. Purdie 1977; Meney et al. 1994), where seedling germination derived from large soil or canopy-stored seed banks is promoted following fire events, but is often restricted to the immediate post-fire year.

For 14 of 24 native species that were observed in the seedling flora at burnt sites, germination appeared to be ongoing (albeit at low density or unsuccessful in some years) rather than episodic. Nonetheless, different species (usually only one or two) dominated the seedling cohort each year. Changes in cohort dominance may be a response to annual changes in seed inputs (Gilfedder & Kirkpatrick 1994; Morgan 1995a), vegetation changes, climatic variability (Rusch & van der Maarel 1992) or chance events (Rusch 1992). A similar pattern of germination has previously been observed for some other non-clonal, large-seeded (≥ 1 mg) temperate grassland species of south-eastern Australia that do not form persistent soil seed banks (e.g. Leucochrysum albicans (Gilfedder & Kirkpatrick 1994) and Rutidosis leptorrhynchoides (Morgan 1995b)). Although data are largely lacking (but see Gilfedder & Kirkpatrick 1994), it is inferred that these particular species are not long-lived (< 10–15 years) because they do not form large rootstocks, nor do they persist in rank swards (Scarlett & Parsons 1990; Gilfedder & Kirkpatrick 1994) and all have a preference for open microsites where recruitment success can be maximized (Gilfedder & Kirkpatrick 1994; Morgan 1997). A strategy of ongoing, non-specialized germination might ensure that seedlings are present to replace individuals of the (few) shorter-lived herbaceous species of this community. Seedling survival may be low in most years for these species, perhaps limited by microsite availability or climatic conditions, but will occasionally be high (Gilfedder & Kirkpatrick 1994; Morgan 1995b) allowing population gains to be made (Warner & Chesson 1985). Leptorhynchos squamatus would appear to be the species in this community that is best defined by such an ecological strategy for population persistence.

Seedling regeneration of native (and most exotic) species was almost non-existent in the long unburnt grassland over the 4 years of this study, as is typical for undisturbed vegetation (Goldberg & Werner 1983; Lunt 1990; Belsky 1992; Morgan 1997, 1998b). The thick litter layer and continuous, dense canopy presumably produced conditions that were unfavourable to the germination of all but a few species (i.e. Critesion murinum and Hypochoeris radicata).

Despite their different long-term fire histories, the grasslands, with the exception of the long unburnt site, do not appear to recruit substantially differently from one another in most years. No significant differences in floristic composition were found amongst all burnt sites on PCA axis 1 over the 4 years of study, although small differences, which were maintained over time, were evident between burnt sites on axes 2 and 3. This is presumably because: (i) few native species recruited in any year, (ii) those that did were common across all sites, (iii) no species had fire-cued germination, and (iv) exotic annual monocots dominate all cohorts in all years. This general result is somewhat surprising given that each of the sites differed substantially at various times in the amount of vegetative cover present at the time of germination, a factor considered by many authors as being an important determinant of germination in grasslands (e.g. Goldberg & Werner 1983; Williams 1992; Thompson et al. 1996; Morgan 1997). The observed germination response may be indicative, however, of the fact that few native species have specialist germination requirements, i.e. ‘gap sensitivity’ is likely to be poorly developed in many species (Morgan 1998b), nor do they have a requirement for fire to release seed from innate dormancy (Morgan 1998d). It may also be that only dense and continuous vegetation, that usually produced by ≥ 5 years’ growth after burning, is sufficient to reduce seedling density and richness over an entire site (McDougall 1989; Lunt 1990; Morgan 1997).

Analyses using PCA followed by repeated measures anova also suggest that the seedling composition of this grassland responds not only to site differences, but also to climatic variation and to significant interactions between the two. Differences in the amount of rainfall, particularly that which falls between April and September when most species germinate, may have significant impacts on the resulting seedling flora as suggested by the co-ordinated changes in all PCA scores of all treatments from 1993 to 1994, in a year of below-average rainfall. Similar effects of climatic variation on seedling floristic composition have more usually been described for annual plant communities (e.g. Gulmon 1992; Guo & Brown 1996) and the results may reflect the dominance of annual species and their sensitivity to environmental variation.

Seedling recruitment appears to play a role in the maintenance of native species richness in this community. Recruitment fluctuates from one year to another, a necessary precursor if it is to contribute to species coexistence (Grubb 1977; Warner & Chesson 1985; Thompson et al. 1996; Kotorova & Leps 1999), and affects all species independently of the effects of management and climate. Coexistence can therefore be fostered by differences in the ‘regeneration niche’ amongst the > 70% of seed-dependent species with similar life-traits. The low levels of recruitment observed in this study, however, suggest that these processes are likely to operate over the temporal scale of decades rather than years, as suggested for short-lived species in limestone grassland communities of north-western Europe (e.g. Rusch & van der Maarel 1992; van der Maarel & Sykes 1993).

The low levels of recruitment of native species suggest that seedlings play a minimal role in the short-term population dynamics of these grasslands. All perennial species are therefore dependent on, and are indeed capable of, vegetative persistence through recurrent fire (Lunt 1990; Morgan 1999). The results predict that relatively long-lived populations with low annual turnover may be common amongst many of the native species, enabling them to remain as a viable constituent of the vegetation even though opportunities for seedling regeneration are relatively infrequent.

A major consequence of the consistently low levels of ‘natural’ recruitment observed in this study is that populations of conservation concern could become quickly locally extinct and permanent change brought upon the community, as has already been documented for a shade-sensitive grassland forb, Rutidosis leptorrhynchoides, following the cessation of burning on railway reserves (Scarlett & Parsons 1990; Morgan 1995b). There will be little potential to restore small-scale species richness once this occurs because very few species form large persistent soil seed banks (Lunt 1997; Morgan 1998a), seedling recruitment is low and difficult to predict, and seedling survival can be exceedingly low (Gilfedder & Kirkpatrick 1994; Morgan 1995b). This may explain why attempts to restore grassland vegetation, either from ‘scratch’ (e.g. McDougall 1989) or in situ, have met with great difficulty.

Management of this fragmented, threatened plant community must therefore consider the persistence of the existing plant pool – the bud- and tuber-bank – as its highest priority. Conditions that promote flowering and, hence, a seed rain must be maintained if seedling recruitment is to occur in a temporally unpredictable manner for species without a persistent seed bank. Perpetuating the long-established management regimes (burning at 1–4 year intervals) is therefore a fundamental minimum management guideline that must be achieved at all sites if the short- to mid-term conservation of the flora is to be achieved. The continuation of monitoring of seedling recruitment is needed to confirm the results of this study, and to ensure that plant populations are in fact being maintained over time. Modifications to existing burning regimes may be warranted, but increasing burning frequency will not necessarily improve seedling recruitment and small-scale patch richness in the short term (Warner & Chesson 1985; Lunt 1990). Rather, it may provide the conditions whereby successful recruitment can ultimately be enhanced (Morgan 1997, 1998b).

Acknowledgements

Tamzin Rollason and Glenn Twenty provided assistance in the field. Ian Lunt, Ken Thompson, Bob Parsons and an anonymous referee provided critical comments on the manuscript and greatly improved its structure. This research was funded by an Australian Postgraduate Research Award.

Received 2 November 1999revision accepted 23 October 2000

Ancillary