• S. K. Florentine,

    Corresponding author
    1. Centre for Environmental Management, School of Science and Engineering, University of Ballarat, Victoria, Australia
    • Centre for Environmental Management, School of Science and Engineering, University of Ballarat, P.O. Box 663, Mt. Helen, Victoria 3350, Australia.
    Search for more papers by this author
  • F. P. Graz,

    1. Centre for Environmental Management, School of Science and Engineering, University of Ballarat, Victoria, Australia
    2. Multidisciplinary Research Centre, University of Namibia, Windhoek, Namibia
    Search for more papers by this author
  • G. Ambrose,

    1. Centre for Environmental Management, School of Science and Engineering, University of Ballarat, Victoria, Australia
    Search for more papers by this author
  • L. O'brien

    1. Centre for Environmental Management, School of Science and Engineering, University of Ballarat, Victoria, Australia
    Search for more papers by this author


Many vegetation restoration works have been undertaken in Australia but only a few of those projects have been assessed for effectiveness. Revisiting earlier restoration attempts and analysing data from them is fundamental to the development of evidence-based prescriptions for future restoration work. Therefore, this study's objectives were to (I) compare plant species composition of different age direct-seeded revegetated sites and (II) determine the effect, if any, of different ages of revegetated sites on the natural recruitment of native plants. The study investigated four fenced restoration sites, dating from 2000, 2001, 2004 and 2005. Results showed that the density of plants surviving varied widely between plots of different ages. The highest density was found in the 2001 plot (2195·.7 stems ha−1), followed by 2000 (1877·8 stems ha−1), 2004 (197·6 stems ha−1) and 2005 (195·4 stems ha−1). An ANOVA showed that the overall amount of seed broadcast does not play a significant (p = 0·437) role in the establishment rate. Overall, Eucalyptus ovata was found to be dominant in the 2000 (794·4 ha−1) and 2001 (971 ha−1) sites. In contrast, Eucalyptus camaldulensis and Eucalyptus viminalis densities were highest in the 2004 (41 ha−1) and 2005 (60·4 ha−1) sites, respectively. PERMANOVA showed highly significant differences in the present plant composition between plots (p < 0·0001), despite similar species mixes used in sowing. Recruitment was not found in any of the sites. The least weed cover and the highest litter cover were found in the 2001 plot. A similar trend was found in the 2000 plot. In contrast, high weed cover and low litter cover were found in the 2004 and 2005 plots. Since one of the major impediments to developing better restoration strategies is the inadequate documentation of past practices, studies such as these may shed some light on how the direct-seeded technique operates in a farm situation. Copyright © 2011 John Wiley & Sons, Ltd.


Ecological restoration is among the most expensive and extensive of natural resource conservation activities worldwide. Over the last two decades the Australian government has spent many millions of dollars on restoration projects nationwide. In 2000–2001 alone, Au$36·4 million was spent to re-establish native vegetation and to provide appropriate habitat for wildlife (Wilkins et al., 2003). Environment Australia (1999) reported that the major on-ground outputs from its Au$27·1 million Bushcare funding programme included 10 000 ha of direct-seeding, the planting of 4·5 million seedlings and erection of 12 000 km of fencing (Wilkins et al., 2003). In addition to this government investment, significant contributions are being made by philanthropic organisations and businesses, while time and material resources are contributed by volunteers towards restoration projects.

Despite substantial financial and human investment in revegetation projects, little or no follow-up monitoring has been undertaken to assess whether the desired ecological benefits have occurred. In a recent review, Brooks and Lake (2007) reported that, of the 2247 restoration projects undertaken in four catchments in Victoria, only 14 per cent included some form of monitoring—mainly photographic records. In most restoration projects the on-ground outcome is therefore largely not evaluated.

A timely review on the state of knowledge of land restoration science in Australia by Lovett et al. (2008) highlights an overwhelming emphasis on faunal assessments—using mainly birds as indicator species—with limited evidence of systematic, quantitative evaluation, particularly of vegetation development. Such assessments are becoming increasingly important, as almost no data are currently available on the effectiveness of past restoration efforts. Michener (1997) had reported that there remains an urgent need for evidence to address the following generic questions: (i) How do we determine whether restoration activities have succeeded or failed? (ii) What can science tell us of the causes of the success or failure? (iii) What should be done in future restoration attempts? (iv) Will new procedures be successful in another environment? (v) What will the new procedures cost? and (vi) Can restoration projects be carried out more economically without prejudicing outcomes? As yet, little seems to have been documented on these issues, and further delay in addressing these questions risks a continuation of practices that do not provide optimal ecological return for the financial and human investment.

Holl et al. (2003) criticised the failure to use large-scale restoration efforts as ‘experiments’, to inform restoration ecologists and on-ground practitioners, as detrimental to the environment. In an era where adaptive management is widely advocated, there is little or no feedback to which operators may respond. Therefore, this project uses the results of underutilised, prior restoration efforts as ‘experimental sites’ in order to gather and synthesise information so as to guide future restoration efforts. Several researchers have indicated that existing restoration sites can be used to assess ecological hypotheses and theories (e.g. Bradshaw, 1987; Harper, 1987), and Michener (1997) stressed that, without proper ecological audit of existing sites, future restoration projects will not benefit from past attempts.

One of the challenging issues in restoration programs is to select an efficient technique for planting (Camargo et al., 2002). Two major types of restoration techniques are currently being used in Australia and other parts of the world: direct-seeding and seedling transplant (Doust et al., 2008; Bonilla-Moheno and Holl, 2009; Bruel et al., 2010). Transplanting of natural regeneration may also be a viable option; however, some logistical and economic complications need to be considered before this technique is used (Engel and Parrotta, 2001). Direct-seeding requires less labour and has higher resource use efficiency than seedling transplanting (Dalton, 1993; Lamb and Gilmour, 2003).

Previously, direct-seeding practices were not favourably considered. This is mainly because several factors hindered the post-establishment of the seedlings (Engel and Parrotta, 2001; Florentine, 2008). These include unfavourable soil conditions (Rasiah et al., 2004), competition from established plants (Putz and Canham, 1992), weeds (Florentine and Westbrooke, 2004) and climatic conditions (De Steven, 1991). In the last few decades, researchers have attempted to address some of these factors through a range of research projects. Studies such as that by Doust et al. (2006) examined suitable microsites for direct-seeding. The authors recorded high establishment rates when the seeds were buried. Similarly Bruel et al. (2010) examined the effect of manual or mechanical sowing techniques and reported that a mechanised planting approach was found to be the best. As indicated previously, finding a suitable time for direct-seeding is also essential. Studies conducted by Bonilla-Moheno and Holl (2009) directly addressed this issue. The authors showed that direct-seeding of mature forest species after natural succession had commenced accelerated the recovery process. Similarly Doust et al. (2008) examined direct-seeding and subsequent establishment of rainforest species in Australia and found that seeded species were more consistent in their establishment and negatively affected by the weeds.

Previous studies have not examined the amount of seeds required for restoration programmes, although the seeds comprise an expensive component of restoration programmes (Broadhurst et al., 2006).

This post hoc study of four direct-seeded patches on an agricultural property was undertaken in an effort to address some of the shortfall in the monitoring of direct-seeding, as suggested by Holl et al. (2003).

It should be noted that the restoration projects considered here were not created as part of a previously designed research project. However, the landholder who established the sites retained a high degree of consistency in techniques over the years, making the sites particularly useful to further enhance our knowledge in this area. In addition, the landholder maintained substantial records of the techniques used, such as the type of chemical used to control weeds, the species he obtained for seeding, as well as the actual dates of the different seeding events. We therefore used the differently aged restoration sites (i) to compare plant species composition in direct-seeded revegetated sites and (ii) to determine the effect, if any, of different ages of revegetated sites on the natural recruitment of native plant species.


Study Site

The study sites consist of four separate, fenced areas of seeded land at Orana, an actively managed grazing property approximately 30 km NW of Ballarat, Victoria, Australia. The soils are primarily of volcanic origin and land use in this area focuses on sheep grazing and crop production. The plots are located on a relatively flat area with an approximately 2–4 degrees south-facing slope. A few old scattered Eucalyptus camaldulensis trees are found in close proximity to the revegetated sites but otherwise there is no nearby remnant vegetation. A mean annual rainfall of 676 mm is recorded by the Australian Bureau of Meteorology (2009) for Trawalla, the closest weather monitoring station. Rainfall in the three months prior to seeding treatments and following the treatments is provided in Table I, as well as the mean maximum temperatures for the seeding month and subsequent month.

Table I. Climatic conditions associated with restoration: rainfall (mm) in the 12 months preceding the seeding of each plot, and in the subsequent 12 months. Data was obtained for Ballarat Aerodrome through the Australian Bureau of Meteorology. Data presented within the bracket is rainfall (mm) in the three months preceding, and following, the seeding of each plot
YearRainfall (mm) in 12 monthsRainfall (mm) July–OctoberMean temperature (°C) October
Prior to seedingAfter seeding
2000803 (344)625 (97)34415·5
2001625 (295)633 (134)29515·2
2004638 (268)658 (175)26817·8
2005658 (294)552 (149)29417·3

Restoration on the property occurred in four phases: 0·18 ha was seeded in 2000, followed by further works in 2001 (0·35 ha), 2004 (0·9 ha) and 2005 (2·4 ha). With the restoration projects the farmer aimed to protect his livestock from wind chill and to create patches of native vegetation as habitat for native animals. All of the revegetation sites were between 10 and 15 m wide and sited along existing fence lines to reduce the cost of fencing and to minimise the loss of productive farmland.

Seed was broadcast during October of each year, using seeds sourced from the Ballarat Regional Seed Bank. The seed had been collected from locations as close as possible to the property, and falling within similar ecological vegetation classes (EVC). Acacia seeds were pre-treated with hot water. Two months prior to seeding, the sites had been prepared by poisoning the existing plant cover using Roundup Powermax® (1·6 L ha−1) and subsequent ploughing (to approximately 30 cm) to reduce competition from pasture and weed species. Seeds of the different species were then mixed and broadcast by hand, but not covered.

Nine species were seeded in three of the four sites, as shown in Table II. Acacia mearnsii was not included in 2004. The Landcare group and Ballarat Seed Bank selected these nine species mainly because they are considered pioneer species and grow well in disturbed areas. Table II provides an estimate of the number of seeds that were broadcast per square metre each year, based on records provided by the property owner. The mean quantities of seed per kilogram, and subsequently per unit area, were calculated from records provided by the Ballarat Regional Seed Bank. It is a standard practice that Ballarat Regional Seed Bank checks the seed percentage germinability. The seeds they supply to restoration programs show close to 95 per cent germination.

Table II. Composition of seed mix used, expressed as number of seeds per m2, during the restoration efforts in the year indicated
SpeciesYear seeds were broadcast
Acacia mearnsii (Black Wattle)17·466·1600·52
A. melanoxylon (Blackwood)14·735·191·570·66
A. pycnantha (Golden Wattle)12·424·381·330·40
A. retinodes (Wirilda)20·077·081·820·75
Allocasuarina verticillata (Drooping Sheoak)24·608·682·620·96
Eucalyptus camaldulensis (River Red Gum)161·2767·3519·012·86
E. ovata (Swamp Gum)64·5115·514·411·98
E. viminalis (Manna Gum)98·0927·8011·044·73
Leptospermum continentale (Prickly Tea-tree)780·86154·4540·7411·46

Data collection

To determine the current species composition and recruitment at each of the sites, five 10 m × 10 m permanent plots were established in each of the revegetation areas and assessed during May and June 2008. All trees within each plot were identified to species level. These species records were used to establish the density of the species within each of the plots. Natural regeneration was assessed by conducting a systematic search for new recruitment of native species (trees or otherwise) within each 10 m × 10 m plot. Total projected canopy cover, percentage of exotic plant cover and litter cover were estimated and recorded for each plot.


Species composition (in stems per hectare) was compared with the original sowing rate (seeds per square metre) using a linear model in R Development Core Team, 2008. Counts of the stems per hectare were log-transformed to achieve normality. Differences in species composition between sites were assessed using the permutational MANOVA [PERMANOVA extension, PRIMER (6·1)®] based on the Bray–Curtis dissimilarity coefficient without transformation. SIMPER (PRIMER 6·1)® was subsequently used to determine the contribution that individual species made towards the mean dissimilarity within each year. An MDS plot was used to assist with the interpretation.



The density of plants surviving in each of the plots was found to differ widely, as shown in Table III. The highest density was recorded in the 2001 plot (2195·7 ha−1), followed by 2000 (1877·8), 2004 (197·6) and 2005 (195·4). Analysis of the data indicated that seed quantity played a significant role in the establishment of the individual species, and accounted for around 29 per cent of the variation in the measured establishment (R2-adj = 0·288, p = 0·001) (Table IV). When differences between years were included in the analysis, the factor was not significant (t = 0·026, p = 0·979), with differences due to year now explaining 46 per cent of the variation (R2-adj = 0·462, p = 0·0002). These differences were primarily attributable to differences in plant survival found for the treatment years 2004 and 2005 (p = 0·003, p < 0·001, respectively). Species-specific differences in establishment could not be determined due to insufficient degrees of freedom.

Table III. Composition of the vegetation within the plots at time of assessment (2008, in stems per hectare). Figures in parentheses indicate the number of dead trees or shrubs detected
SpeciesYear seeds were broadcast
  1. NB, Not broadcast.

Acacia mearnsii (Black Wattle)477·8128·6NB10·0
A. melanoxylon (Blackwood)100·088·622·214·6
A. pycnantha (Golden Wattle)33·3657·130·020·8
A. retinodes (Wirilda)794·4971·433·338·3
Allocasuarina verticillata (Drooping Sheoak)72·241·012·21·7
Eucalyptus camaldulensis (River Red Gum)029·041·134·2
E. ovata (Swamp Gum)277·871·428·914·6
E. viminalis (Manna Gum)105·6208·627·860·4
Leptospermum continentale (Prickly Tea-tree)16·702·00·8
Acacia unidentified(33·3)(134·3)(3·0)
Eucalyptus unidentified(538·9)(1142·9)(4·0)(0·4)
Total (stems per hectare)2450·03472·9204·6195·8
Total live1877·82195·7197·6195·4
Table IV. ANOVA resulting from the fitting of the linear model of stems per hectare against seeds per m2 and year of treatment
CoefficientsEstimatet valuep value
Intercept5·372 (± 0·54)9·8554·300
Seed density−0·008 (± 0·01)−0·7890·437
Year 2001−0·300 (± 0·30)−0·5250·603
Year 2004−2·049 (± 2·05)−0·6390·003
Year 2005−0·254 (± 2·54)−0·6840·000

Species Composition

Acacia retinodes was found to be dominant in the 2000 (794·4 stems ha−1) and 2001 sites (971 stems ha−1). In contrast, Eucalyptus camaldulensis and Eucalyptus viminalis densities were the most abundant species in the 2004 (41 stems ha−1) and 2005 (60·4 stems ha−1) sites, respectively (Table III).

The highest percentage of dead stems was recorded in the 2001 plot (36 per cent) followed by 2000 (23 per cent), 2004 (3 per cent) and 2005 (less than 1 per cent). No native plant recruitment was found in any of the plots we assessed.

The PERMANOVA showed highly significant differences in the present plant composition between plots (F = 18·166, p < 0·0001). Excluding the dead trees from the analysis did not change the conclusion (F = 14·699, p < 0·0001).

The MDS plots provided in Figure 1 (a) and (b) show that the composition of the vegetation within most of the seeded plots was relatively consistent, though the (within-plot) variation in composition seems to be higher in the plot seeded in 2004. This is supported by the SIMPER analysis, which showed a within-plot similarity of above 70 per cent for the plots established in 2000, 2001 and 2005. In the plot established in 2004, the clumping is not as tight and within-plot similarity is 57 per cent.

Figure 1.

MDS plots of the species composition within each of the five sites in all four years, including (a) all trees and (b) trees alive at the time of the enumeration, with stress values of 0·08 and 0·1, respectively [2000 (▴), 2001(▾), 2004(▪), 2005(♦)]. This figure is available in colour online at

The SIMPER analysis determined the species that contributed most to the dissimilarity in species composition between the different years. This is presented in Table V. The values in parentheses represent the percentage contribution to the dissimilarity. It is evident that Acacia retinodes is the source of most dissimilarity between plots. The same species also created most of the variations within the 2000, 2001, 2005 plots, though, somewhat surprisingly, not in the 2004 plot.

Table V. Species—live plants only—contributing to dissimilarity between plots (per cent), seeded in the years indicated
  Year of seeding
Year of seeding2001A. pycnantha (36.7)
A. retinodes (32.3)
2004A. retinodes (35.4)A. retinodes (44.8)
A. mearnsii (20.8)A. pycnantha (29.3)
2005E. viminalis (29.7)A. retinodes (38.5)E. viminalis (37.3)
E. camaldulensis (17.5)A. pycnantha (27.45)A. retinodes (18.0)

The highest canopy cover was found in the 2001 direct-seeded site, followed by 2005, 2004 and 2000. The 2004 seeded plot showed very high weed cover (80 per cent) in comparison to other plots. Less weed cover and the highest litter cover were found in the 2001 plot. A similar trend was found in the 2000 plot. In contrast, high weed cover and low litter cover were found in the 2004 and 2005 plots (Figure 2).

Figure 2.

Mean canopy, weed and litter cover of plots when assessed in 2008.


This paper (I) compares plant species composition of different aged direct-seeded revegetated sites and (II) determines the effect, if any, of different ages of revegetated sites (or seedling establishment in different years) on the natural recruitment of native plant species.

The use of existing revegetation sites imposed some constraints on the analysis. While the differently aged revegetated sites were direct-seeded and shared similar soil types and topography, each site differed in size and possibly concomitant edge effects. In addition, restoration actions were not replicated to detect within-year variability, since the seeding endeavours were not part of formalised experimental trials. This lack of replication precludes the detection of statistically significant interactions between species. Finally, external influences, such as the occurrence of seed predators at the time of sowing, were not monitored. Therefore, extrapolations from this study should be undertaken with caution.

Our results show that there was little agreement in the final species composition between the restoration attempts, despite the similarity in the seed source. Much of the dissimilarity between the various plots was due to differences in the occurrence of Acacia pycnantha and A. retinodes. Therefore, it is instructive to consider these two species more closely. Table III shows that the two species follow different patterns of survival in the 2000 and 2001 plots. This should be considered in the context of the growth requirements of the species. As we pointed out in Table VI, while Acacia pycnantha prefers well drained soil types, A. retinodes is more often found on poorly drained soils (Costermans, 2005). A corresponding pattern may be observed for the two species in the plot established in 2005, though to a lesser extent. E. viminalis has a similar preference for well drained soils, and its establishment pattern would be expected to follow that of Acacia pycnantha.

Table VI. Characteristics of seed used in these restoration programs, with their requirements for germination and subsequent growth features
SpeciesSeed typePre-treatmentGerminates (burial)Establishment requirementsInitial growth
  • Seed type: large/ExtDorm (thick testa): large seed with external dormancy (impervious testa). Small/LDom: small seed with little dormancy. Fine/NoDorm: fine seed with no dormancy.

  • Pre-treatment: BW + SO/SC: boiling water then soaking for 24 h, or scarification, +/ − Strat: germination rate improved by stratification, +/− Smk: germination rate improved by smoke treatment: no pre-treatment required.

  • Germinates (burial): readily: germinates readily, xy d: germinates in x to y days. (Germination is slower than this in cold conditions.), Shallow: seeds should not be deeply buried; may fail on surface that becomes dry, Deep: seeds can be buried in up to several centimetres of soil.

  • Establishment requirements: (loam): prefers loamy soils, (Alluvial): prefers alluvial soils. WD: well-drained soil, WL: can tolerate periods of waterlogging, DRY: can tolerate extended dry periods, SA: sand, CL: clay, AD: adaptable. Initial growth: rapid or very rapid.

  • a

    Dormancy depends on seed coat hardness. Softer seeds harden and germinate later.

  • b

    Gray and Knight (eds) (1993).

  • c

    Elliot and Jones (1982).

  • d

    Elliot and Jones (1980).

  • e

    Royal Botanical Gardens, Sydney (2010).

  • f

    Boland (1997).

  • g

    Elliot and Jones (1993).

  • h

    Charles Sturt University Herbarium (2010).

  • i

    Elliot and Jones (1986).

  • j

    Australian National Botanic Gardens (2008).

  • k

    Merritt et al. (2007).

  • l

    Banyule City Council (2007).

Acacia mearnsiib,c,d,e,f,lLarge/ExtDorm (thick testaa)BW + SO/SCReadily, 7–12 d (shallo–deep)FrRes WD,DRYVery rapid
A. melanoxylonb,c,d,e,f,h,lLarge/ExtDorm (thick testaa)BW + SO/SCReadily, 7–12 d (shallow–deep)(Moist) ,∼FrTend WD,WL,DRY,ADRapid
A. pycnanthab,c,d,e,f,hLarge/ExtDorm (thick testaa)BW + SO/SCReadily, 7–12 d (shallow–deep)FrRe WD,DRY,ADRapid
A. retinodesb,c,d,e,f,jLarge/ExtDorm (thick testaa)BW + SO/SCReadily, 7–12 d (shallow–deepFrRes AD, can suckerRapid
Allocasuarina verticillatac,kSmall/LDorm(+/− Strat)Readily, 15–30 d (shallow)FrRes (WL),CL,SA,ADRapid
Eucalyptus camaldulensisd,f,iFine/NoDormReadily, 2–12 d (shallow)(Loam) FrRes WL,DRY,AD,SAVery rapid
E. ovatab,d,f,iFine/NoDormReadily, 2–12 d (shallow)(Moist) FrRes WL,WD,(DRY),CLRapid
E. viminalisb,d,f,iFine/NoDormReadily, 2–12 d (shallow)(Moist alluv) FrRes WD,ADRapid
Leptospermum continentalec,g,h,lFine/LDorm(+/− Smk)Readily, 3–5 wks + d (shallow)(Moist) FrResist. WL,SA,ADRapid

Both Acacia and Eucalyptus species occur in high densities on all of the revegetated sites. It is anticipated that the densities of these species will decline during subsequent years due to self-thinning. The current densities of species such as Allocasuarina verticillata and Leptospermum continentale are not high, in particular that of L. continentale. These results need special attention here. Previous studies (Battaglia and Reid, 1993; Doust et al., 2008) have shown that larger seeded species have a tendency to establish in higher densities than smaller seeded species. This tendency may have been enhanced because Acacia seeds received pre-treatment, possibly increasing the germination rate of these species (Table VI).

However, the Eucalyptus species used in this study also have smaller seeds, so we should ask why their current density is significantly higher than that of Allocasuarina verticillata and Leptospermum continentale. In general, we propose that, because nearly all chosen eucalypts are adaptable to a range of germination conditions (Table VI), they are likely to find suitable temperature and soil moisture conditions for germination and, subsequently, establishment (Battaglia and Reid, 1993).

The small-seeded species A. verticillata and L. continentale did not establish well across the restoration sites. It may be possible that during the seed broadcast, seeds may have landed outside the revegetation area or that seed predators may have later removed the seeds (Hughes and Westoby, 1990, Garcia-Orth and Martinez-Ramos, 2008). Also, it is very well documented that the litter layer functions as a physical barrier to the establishment of small-seeded species (Woods and Elliott, 2004). As we described above, the sites had been prepared for revegetation by poisoning the existing plant cover. The dead and dried remnant materials may subsequently have prevented A. verticillata and L. continentale seed from reaching the soil.

Finally, the two species may have germinated but failed to establish during subsequent months. Without post-seeding monitoring data, it is necessary to speculate. It seems to us that combination of above-mentioned reasons may have played a critical role in preventing the germination and subsequent establishment of these seeds. Studies conducted by Turner (2001) and also Doust et al. (2006) found that smaller-seeded species were very poor in their germination rates and subsequent establishment rates. They suggested that it is essential to consider these findings for future restoration programs, particularly when aiming to reduce the cost of the restoration program and also seed wastage. As we pointed out in the methods, the seeds of these nine species were broadcast, but not covered. We believe this aspect deserves some attention. Doust et al. (2006) show that a higher germination rate was observed in buried seeds than seeds that had been broadcast. A. verticillata and L. continentale should perform better if the seeds are shallow-buried.

The second objective of this study was to detect whether any of the different-aged revegetation plots resulted in natural recruitment from the sown trees and shrubs. This is important if restoration efforts aim to establish self-sustaining ecosystems, in line with the goal of the Society for Ecological Restoration International, as adopted by the International Union for the Conservation of Nature (IUCN) (SER, 2004). The results show, however, a complete absence of recruitment—even though some species such as Acacia pycnantha and Eucalyptus camaldulensis in the oldest direct-seeded site (just over eight years-old) showed signs of old capsules and flower buds on the ground. We anticipate that they may continue to supply a significant amount of seed every flowering season. It may therefore be speculated that at least some of these trees have supplied seeds to the sites and seeds of Acacia species may have stored in the soil. However, lack of recruitment in the study sites deserves some attention in this discussion. There are several biotic and abitic factors that might be involved in this lack of recruitment. The first is seedbed conditions. For instance, seed germination and seedling establishment may have been adversely affected by a non-receptive seedbed (Florentine and Westbrooke, 2004). Also, lack of seedling establishment, attributed to the thick grass cover and leaf litter present across the treatment, may have prevented the seed from reaching the soil (Aide et al., 1995). Further, Whelan et al., (1991) found that seed predation can eliminate significant numbers of seeds within a few weeks of seed arrival, although this possibility would require further study. Even if the seeds reach the soil substrate they face competition with the already established ground cover species (Parrotta, 1992). Reiners et al. (1994) demonstrated that continuous use of land for pasture and subsequent land degradation may delay natural recovery. We anticipate that (future) continuous accumulation of litter will change this condition and create more suitable conditions for the soil-stored seeds to germinate.

Our results lead to following main conclusions. The density of plants surviving varied widely between plots of different ages, most likely reflecting variations between years in the conditions required for germination and seed survival. While the overall amount of seed broadcast does play a role in the establishment rate of the selected species, this effect may be overshadowed by annual variations in environmental conditions that affect germination or establishment. In this study, A. verticillata and L. continentale did not establish well across the sites. We suggest three possible reasons: (i) these species seeds may not have germinated after broadcast, (ii) seeds may have been predated, (iii) seeds may have germinated but a subsequent dry spell may have killed the seedlings before they developed large root systems. Since we lack data on the post-sowing conditions, it is difficult for us to suggest means of avoiding this outcome. Further research should be conducted on this matter. At this stage we can recommend increasing the quantity of seed broadcast for these two species, ensuring barriers to seed penetration such as leaf litter or hard soil crusts are minimised and attempting to sow when soil surface moisture is likely to be maintained (i.e. before protracted dry periods are likely).

It should be noted that planting native species as a windbreak in farmlands is much better than planting a single row of exotic species. Whichever species are used, the successful establishment of target species does not necessarily imply that the revegetation has achieved long-term success. This will depend on the ability of plants to self-propagate under the prevailing environmental conditions. Further work on the species composition of the soil seed bank and an avifauna utilisation study on revegetated plots may provide further information about the effectiveness of these restoration endeavours.


Authors thank Robert Dunn, who generously allowed us the use of his property and supplied information about the site and plots. They also thank Michelle Butler (Department of Primary Industries, Victoria), who provided financial support for this study. Authors are grateful to Arnaud Fourrier for help in the field. Finally they thank two anonymous reviewers for providing valuable comments and suggestions on earlier versions of the manuscript.