Adaptive restoration of sand-mined areas for biological conservation



    Corresponding author
    1. Ecosystem Management, School of Environmental Sciences and Natural Resources Management, University of New England, Armidale NSW 2351, Australia;
      *Present address and correspondence: Jason Cummings, School for Field Studies, Centre for Rainforest Studies, PO Box 141, Yungaburra QLD, 4884 Australia (fax + 612 67732769; e-mail
    Search for more papers by this author

    1. Ecosystem Management, School of Environmental Sciences and Natural Resources Management, University of New England, Armidale NSW 2351, Australia;
    Search for more papers by this author

    1. Statistics, School of Mathematics, Statistics and Computer Science, University of New England, Armidale NSW 2351, Australia; and
    Search for more papers by this author

    1. Alcoa World Alumina Australia, Applecross WA 6953, Australia
    Search for more papers by this author

*Present address and correspondence: Jason Cummings, School for Field Studies, Centre for Rainforest Studies, PO Box 141, Yungaburra QLD, 4884 Australia (fax + 612 67732769; e-mail


  • 1Adaptive management approaches to ecological restoration are current best practice. The usefulness of such an approach was tested in this study by implementing repeated experiments that examined restoration options for derelict sand mine sites dominated by Imperata cylindrica. Reclamation of degraded land that is dominated by I. cylindrica is a common problem throughout the tropics.
  • 2Initially, the hypothesized barrier to regeneration was limited seedling establishment because of I. cylindrica competition. After burning the grassland, woody weed control and planting of seedlings were implemented in factorial combination.
  • 3Seedling survival 28 months after planting averaged 26%, with < 1% of all seedlings establishing to a height > 1 m. The hypothesis that a transition barrier comprising solely biotic interactions restricted regeneration of native woody cover was rejected after seedlings and natural regeneration failed to thrive in this experiment.
  • 4A revised hypothesis, that the transition barrier comprised a combination of abiotic limitations (soil deficiencies) and biotic interactions (Wallabia bicolor browsing and I. cylindrica competition), was developed. A second experiment tested this hypothesis by removing W. bicolor (fencing), slashing the I. cylindrica, adding organic mulch and planting a mixture of native pioneer and secondary successional woody species in factorial combination.
  • 5Seedling survival was 61% in the second experiment and mulching significantly enhanced the survival and growth of all planted species. Planting alone reduced the regeneration of I. cylindrica after slashing. Native woody cover establishment was maximized by planting seedlings in mulched treatments.
  • 6Synthesis and applications. Taken together, these experiments support the hypothesis that there is a barrier restricting regeneration of native woody cover, and the barrier probably comprises both abiotic and biotic components. By adopting an adaptive management approach to the ecological restoration of sites, significant insights into their management requirements have been gained, supporting the current best practice restoration framework. Insights gained through monitoring and adaptation will be used to update the reserve plan of management, enhancing restoration of this severely degraded area and promoting connectivity of native woody cover within the conservation estate.


Mineral sands mining occurs in coastal southern Africa, India, North America and Australia (Lubke & Avis 1998). After mineral sands mining, soil organic matter, seed bank and nutrients are diluted or historically absent. Disruption of the soil profile results in leaching of nutrients and moisture loss (Prosser & Roseby 1995). While restoration techniques for current mining operations have improved considerably in comparison with historical operations (Bell 2001), viable restoration techniques for derelict mines, where fresh topsoil is not available, have not been documented.

Mineral sands mining occurred along the Bongil Peninsula on the north coast of New South Wales (NSW, Australia) from the mid-1960s until the early to mid-1970s. Attempts were made in the mid-1970s to establish flooded gum Eucalyptus grandis W. Hill ex Maiden (plant nomenclature throughout follows Harden 1990–2002) plantations directly onto sand mine tailings. In several patches on the peninsula, plantation tree establishment failed. With occasional fires caused by recreational users of these sites, blady grass Imperata cylindrica P. Beauv. became dominant in open areas of what is now Bongil Bongil National Park (BBNP). Imperata cylindrica is a tufted perennial grass that spreads by rhizomes. It is native to the east coast of NSW and is commonly found in fire-prone areas (Harden 1990–2002) on a variety of soil types, including highly leached soils with low pH, fertility and organic matter (Dozier et al. 1998). Imperata cylindrica is a weed throughout the tropics and subtropics, its rhizomes forming a dense mat that can exclude most other vegetation (Dozier et al. 1998).

Reclamation of land dominated by I. cylindrica is a crucial environmental and forestry issue in south-east Asia (Kuusipalo et al. 1995). At rehabilitated sand mine sites on the NSW north coast, increases in introduced species and altered native species composition and vegetation structure have been recorded (Buckney & Morrison 1992; Fox et al. 1996). Absence of wet sclerophyll species was attributed to the xeric condition of rehabilitated sites, which was related to the disrupted soil profile. Various management techniques for I. cylindrica-dominated sites have been trialed. Physical and chemical control of I. cylindrica reduce cover for 12–24 months (Dozier et al. 1998; Ramsey et al. 2003). Long-term control requires the establishment of competitive species (Kuusipalo et al. 1995; Dozier et al. 1998), otherwise reshooting from rhizomes allows re-infestation (Ramsey et al. 2003). Establishing desirable native woody cover at the I. cylindrica-dominated sites in BBNP is imperative to initiating a self-sustaining restoration process and is a short-term priority of the NSW National Parks and Wildlife Service (NPWS; National Parks and Wildlife Service 1999).

The promotion of vegetation succession as a means of restoring areas dominated by I. cylindrica lends itself to the adoption of an ecological state-and-transition model for understanding and managing the system (sensu Westoby, Walker & Noy-Meir 1989). Imperative to the success of this approach is the accurate identification and management of abiotic limitations or biotic interactions that restrict transitions between states (Hobbs & Norton 1996; Whisenant 1999). The ecological restoration of I. cylindrica-dominated derelict sand mines in BBNP presented an opportunity to test the usefulness of the current best practice ecological restoration framework (Hobbs & Harris 2001; Lake 2001) while improving our understanding of the system and enabling the development of a state-and-transition model. The aims of this study were to: (i) identify barriers limiting ecological restoration of I. cylindrica-dominated sand-mined sites; (ii) recommend management techniques to initiate succession towards vegetation communities more appropriate for conservation; and (iii) examine the usefulness of adaptive management in restoring highly degraded land in a conservation estate.

Materials and methods


BBNP is located on the mid-north coast of NSW (10 km south of Coffs Harbour, Australia). Neighbouring land uses include forestry, grazing, rural–residential and low-density urbanization (National Parks and Wildlife Service 1999). Subtropical Coffs Harbour has an average daily minimum and maximum temperature of 14·0 °C and 23·2 °C, respectively (1943–2001; Bureau of Meteorology 2002). Annual rainfall averages 1708 mm, with a maximum in March and minimum in September. Conversely, evaporation peaks in December–January, with a maximum of 202 mm month−1 and minimum in June of 78 mm month−1. The climate during the study was abnormal; from August 1999 to August 2002, Coffs Harbour received less than the long-term average rainfall in 26 of 37 months. Over this period, evaporation was greater than precipitation in 22 months.

Mineral sands mining of the Bongil Peninsula sand mass involved clear felling of vegetation along the mining path. Plantations of E. grandis and sydney blue gum Eucalyptus saligna Sm. were planted on sand mine tailings and were left unmanaged. Volunteer I. cylindrica grasslands established where plantations failed. Projected foliage cover of I. cylindrica at these sites averaged 39·0 ± 1·5% (mean ± 1 SE, this study), with emergent clumps of the exotic woody weeds, lantana Lantana camara L. and bitou bush Chrysanthemoides monilifera (L.) Norlindh. Occasional native Acacia spp., black sheoak Allocasuarina littoralis (Salisb.) L. Johnson, coast banksia Banksia integrifolia ssp. integrifolia L. f. and geebung Persoonia stradbrokensis Domin also established. Canopy tree cover of these native species and infrequent plantation trees in the I. cylindrica grassland averaged 15·7 ± 2·1%. Based on aerial photograph interpretation, a continuous canopy of mixed hardwood and rainforest vegetation existed before sand mining. Research sites were established within the I. cylindrica-dominated grasslands on flat sandy ground, where sand mining had occurred and plantation establishment had failed.

experiment 1: vegetation management alone

A randomized block experiment was established to contrast the effectiveness of planting primarily woody species of different community types (wet sclerophyll and rainforest) and post-planting woody weed control (in factorial combination) in restoration of the blady grass area. In April 2000 three replicate 50 × 100-m plots were established; each plot was burnt to suppress temporarily the biomass of I. cylindrica and aid in the control of L. camara and C. monilifera to allow planting, and to initiate germination of desirable species that might be present in the seed bank. Remaining woody weed clumps received Roundup® (Glyphosate 1 : 100 dilution; Scotts Company, Marysville, Ohio, USA) applied by hand. Infrequent, naturally occurring trees or shrubs (above) and occasional plantation trees were not removed.

Six 25 × 25-m subplots were established at each plot and were randomly allocated to one of the following treatments: (i) no further management; (ii) annual woody weed control; (iii) planting rainforest seedlings with annual woody weed control; (iv) planting rainforest seedlings; (v) planting wet sclerophyll seedlings with annual woody weed control; and (vi) planting wet sclerophyll seedlings. An additional three control subplots were established to measure background variation in response variables near the treatment plots. Planting was conducted after any remaining L. camara and C. monilifera plants were physically removed. Within each planted subplot, five seedlings of each of 14 species of local wet sclerophyll or rainforest communities were planted (depending on species, these were 50–100 cm tall). In November 2000 and January 2002, experimental post-planting weed control was conducted, whereby C. monilifera and L. camara were hand-removed if close to native woody species or otherwise sprayed with Roundup® (1 : 100 dilution).

Measurement of variables reflecting ecosystem function and regenerating plant composition occurred before (August 1999) and after (August 2001) treatment implementation at the subplot level. Vegetation species composition and cover were recorded in six 2 × 2-m quadrats in each subplot. Plant species cover data were summed in three groups: understorey woody natives, herbaceous natives and introduced herbaceous species. Each variable was averaged across quadrats to give a single subplot value for analyses. Four surface soil (0–50-mm) samples from quadrats within each subplot were bulked and analysed for organic matter, pH (water), electrical conductivity, total nitrogen, nitrate, ammonium, sodium, potassium, calcium and magnesium content (following methods from Rayment & Higginson 1992). Survival and arbitrarily selected height classes (< 0·5 m, 0·5–1·0 m, > 1·0 m) were recorded for each planted seedling in November 2000 and October 2002.

Planted seedling survival and abundance within height classes were compared between community type in November 2000 (6 months post-planting) and between community type and weed control treatment in September 2002 (28 months post-planting) using generalized linear models (binomial error distribution) and analysis of deviance. Soil chemical variables and regeneration cover (2001) were compared using the pretreatment values (1999) as a covariate. Planting did not affect regeneration and was removed from the model, allowing the weed control treatments to be pooled across planting treatments. A mixed model was then used to account for the unbalanced design (n= 9 for both post-planting weed control and no further management) and analyse the effects of post-planting weed control versus no further management and the experimental controls (n = 3; using the plot by weed control interaction as a random effect).

experiment 2: vegetation and soil management

A split-plot factorial experiment was implemented across three replicate 25 × 25-m sites in April 2001. Lantana camara and C. monilifera were physically removed before sites were slashed and fenced to a height of 1·2 m. Mulching and regenerating woody weed control applications were applied in factorial combination. Two 15 × 10-m plots per site were mulched with community green waste to a depth of 50 mm, and another two 15 × 10-m plots were not mulched. One of each of the mulched and non-mulched plots had chemical weed control (Roundup® 1 : 100) applied 8 months after initial treatments were implemented, resulting in four plot treatments: (i) mulch, weed control; (ii) mulch, no weed control; (iii) no mulch, weed control; and (iv) no mulch, no weed control. Plots were split into six establishment subplots of 5 × 5 m that were treated by: (i) planting four species of tubestock at 1 seedling m−2; (ii) applying DAP fertilizer (Hi-Fert, Melbourne, Australia) at 100 kg ha−1; (iii) planting four species of tubestock at 1 seedling m−2 with DAP fertilizer applied at 100 kg ha−1; (iv) planting four species of tubestock at 2 seedlings m−2; (v) hand-broadcasting heat-treated coastal wattle Acacia longifolia ssp. sophorae (Labill.) Court seed at 500 seeds per 25-m2 subplot; and (vi) no further management. Four species (A. l. sophorae, B. i. integrifolia, swamp oak Casuarina glauca Sieber ex Sprengel and lilly pilly Acmena smithii[Poir.] Merr. and L. M. Perry) were selected for planting treatments based on their occasional presence in nearby and similarly degraded vegetation and the availability of seedlings of local provenance.

To assess the immediate influence of treatments on soil chemistry, surface soil samples were sampled, from each subplot, 4 and 11 months after treatment implementation to a depth of 5 cm, in August 2001 and March 2002. In mulched treatments, samples were taken from underneath the 5-cm mulch layer. Samples were analysed for the same variables as in experiment 1, except soil nitrate, and with the inclusion of total and available phosphorus and soil moisture content (following methods from Rayment & Higginson 1992). Chemical response variables were analysed using a linear mixed effects model with a normal error distribution (eqn 1). The model used was of the form:

image(eqn 1)

where Y was the observed value in a subplot ijklr, B accounts for differences between sites (i = 3), M refers to the main plot mulching treatments (j = 4), E refers to subplot establishment treatments with six levels (k; see above) and T refers to sampling time with two levels (l = 2). To account for the split-plot design, the site by plot (BM) and site by plot by subplot (BME) interaction terms were modelled as random effects (as was the residual error e). The split-plot design accounts for repeated measures when only two times are analysed. The remaining terms were modelled as fixed effects.

In December 2002, an assessment of seedling survival, health and growth was conducted. For each planted seedling, survival, height, canopy width and basal girth 50 mm above ground was recorded. The health of each surviving seedling was rated as either healthy or unhealthy; healthy individuals displayed no signs of chlorosis, moisture stress, browsing damage or fungal presence across approximately 90% of their canopy. Survival and health were analysed, for each species separately and combined, using a general linear mixed model with a binomial error distribution (Equation 2). The model used was of the form:

image(eqn 2)

where p was the observed proportion of individuals in a given class in a subplot ijkr, and E has three levels (see above; k= 3). Each of the remaining terms in equation 2 was identical to those described for equation 1, without the temporal element (T or its associated terms). The analysis of variance in the glmmPQL function (R MASS library) accounts for overdispersion, common in binomial distributions, thus analysis of deviance was not required (Venables & Ripley 2002). Seedling canopy width, height and basal girth of surviving seedlings were analysed using equation 2, but as continuous response variables (Y, as per equation 1) with normal error distributions.

Understorey canopy development and natural regeneration were also sampled in December 2002. Three 1 × 1-m quadrats were haphazardly thrown within each subplot and the composition and estimated projected cover of naturally regenerating and planted species was recorded. Species were allocated to one of four functional groups for analyses: introduced herbaceous, introduced woody, native herbaceous and native woody. Total plant cover was the sum of the cover of each functional group plus that afforded by planted seedlings. Imperata cylindrica cover was analysed as part of native herbaceous cover and separately, as it dominated this group. Within each subplot, canopy cover was measured 30 cm above the ground and at breast height using a spherical densiometer (Forest Densiometers, Bartlesville, Oklahoma, USA), the difference being analysed as the subcanopy cover established by the restoration treatments. Natural regeneration and subcanopy cover in subplots were analysed using equation 1, without the temporal element (T or its associated terms).

For all models the restricted maximum likelihood algorithm was used to estimate variance parameters (Patterson & Thompson 1971; Gilmour et al. 2002). To ensure that the assumptions of normality and even dispersion were not violated, diagnostic plots were constructed: residuals vs. fitted values were plotted for dispersion and residuals vs. standard normal quantiles for normality (Quinn & Keough 2002). For analyses with a normal error distribution, natural log or square root transformations were often required. Therefore, predicted means (Johnson & Omland 2004) are usually presented with 95% confidence intervals, back-transformed as required. The MASS (Venables & Ripley 2002) and nlme (Pinheiro et al. 2003) libraries were used within the statistical program R for all analyses (Ihaka & Gentleman 1996). For all models, planned contrasts identified treatment differences where main effects were significant (P = 0·05). Throughout CL refers to the range of lower and upper 95% confidence limits.


experiment 1: vegetation management alone

Overall seedling survival averaged 67% and 26% at 6 and 12 months after planting, respectively. Browsing of rainforest and wet sclerophyll seedlings by swamp wallabies Wallabia bicolor Desmarest was evident. No differences were found between survival of seedlings from wet sclerophyll or rainforest communities at 6 and 28 months after planting. A greater proportion of surviving individuals from the wet sclerophyll community had reached and maintained a height > 50 cm compared with rainforest seedlings, at 6 (χ2 = 38·5, d.f. = 1, P < 0·001) and 28 months (χ2 = 47·7, d.f. = 1, P < 0·001) after planting (Table 1). There was no effect of post-planting weed control on either survival or growth at 28 months. There was no effect of treatments on any soil chemical variable.

Table 1.  Survival and proportion of surviving seedlings in different height classes in the first experiment
Post-planting (months)Community typeSurvival (%)0–50 cm (%)50–100 cm (%)> 100 cm (%)
  1. Predicted mean with 95% lower and upper confidence limits in parentheses; means in the same column at the same time with different superscripts were statistically different (P = 0·05).

6Rainforest72·6 (67·0, 77·6) 99·2 (96·6, 99·8) 0·8a (0·2, 2·6)0a
Wet sclerophyll63·4 (57·5, 69·0) 85·0 (78·9, 89·7)13·1b (9·4, 18·0)0·4b (0·0, 30·9)
28Rainforest23·4 (18·9, 29·0)100 0a0
Wet sclerophyll27·7 (22·7, 33·3) 61·2 (48·7, 72·3)36·0b (27·0, 46·1)0·5 (0·0, 92·5)

A total of 68 plant species was identified to at least genus over 2 years in the first experiment. Six of the 10 most frequently encountered species were introduced. Treatments did not influence native or introduced herbaceous cover, which averaged 60% (CL 49·9–69·1) and 5% (CL 1·4–9·4), respectively, across subplots in 2001 (Table 2). Of the native herbaceous cover, I. cylindrica contributed 39% in 1999, increasing to 46% in 2001. Ground preparation reduced introduced woody cover by 84% (F = 30·7, d.f. = 1, 5, P= 0·003) and native woody cover by 19% (F = 22·3, d.f. = 1, 5, P= 0·005) compared with controls (Table 2).

Table 2.  Regenerating plant cover of different functional groups in response to weed control treatments from the first experiment in 2001
 Introduced herbaceous (%)Introduced woody (%)Native herbaceous (%)Native woody (%)
  1. Predicted mean with 95% lower and upper confidence limits in parentheses; means in the same column with different superscripts were statistically different.

  2. NS, not significant.

Experimental control1·37 (0·0, 40·3)16·7a (5·5, 42·2)79·9 (38·9, 159·1)19·5a (8·8, 30·2)
No further management8·14 (4·2, 12·1) 3·1b (1·3, 3·6)50·7 (40·7, 60·8) 3·9b (0·1, 7·6)
Post-planting weed control9·9 (5·4, 14·6) 2·2b (0·6, 2·3)52·5 (41·8, 63·5) 3·6b (0·0, 7·4)
StatisticNSF = 17·0, d.f. = 2, 4, P= 0·01NSF = 10·4, d.f. = 2, 4, P= 0·03

experiment 2: vegetation and soil management

In the 11 months following implementation, few marked effects of treatments were detected on soil chemistry, with comparable variables being similar between experiments (Table 3). Mulch application increased pH slightly, from 5·4 in both non-mulched (CL 5·3–5·6) treatments to 5·6 and 5·7 (CL 5·5–5·8) in mulched treatments (F = 18·9, d.f. = 1, 8, P = 0·003). Soil moisture content was affected by subplot treatments (F = 2·7, d.f. = 5, 40, P= 0·04) irrespective of mulch application. Planting of seedlings reduced available soil moisture, from 0·4%, 0·3% and 0·3% in unplanted subplots (CL 0·2–0·6) to 0·2% in each planted subplot (CL 0·1–0·3; F = 11·6, d.f. = 1, 56, P = 0·001). Subplot treatment effects were detected for both total and available phosphorus (F = 3·2, d.f. = 5, 40, P= 0·02; F= 15·0, d.f. = 5, 40, P < 0·001, respectively), with fertilizer application increasing soil total phosphorus from 94·2 p.p.m. (CL 78·8–109·7) to 106·9 p.p.m. (CL 95·4–118·4; F = 11·4, d.f. = 1, 56, P= 0·001) and available phosphorus from 4·4 p.p.m. (CL 3·2–5·6) to 9·1 p.p.m. (CL 7·3–10·9; F = 83·9, d.f. = 1, 56, P < 0·001). Mulch application increased potassium in the sands to detectable levels, from 0·0 meq 100 g−1 (CL 0·0) in non-mulched plots to 0·1 meq 100 g−1 (CL 0·0–0·1) in plots with mulch (F = 27·1, d.f. = 1, 8, P < 0·001).

Table 3.  Soil chemical variables of I. cylindrica-dominated sites in the first and second experiments
Soil chemical variableExperiment 1 (1999–2001)Experiment 2 (2001–02)
  • Means are presented with upper and lower 95% confidence limits in parentheses.

  • NS, not sampled.

  • *

    This value was affected by treatments implemented in the second experiment, as referred to in the text.

Organic matter (%)  1·6 (1·4, 1·8)  1·7 (1·6, 1·9)
Soil moisture content (%)NS  0·2* (0·2, 0·3)
pH H2O  5·7 (5·5, 5·9)  5·5* (5·5, 5·6)
Conductivity (µS cm−1) 16·6 (14·2, 18·8) 26·5 (25·0, 28·1)
Total nitrogen (p.p.m.)502·8 (411·9, 593·7)338·6 (284·7, 397·2)
Nitrate (p.p.m.)  3·0 (1·1, 4·9)NS
Ammonium (p.p.m.)  6·5 (3·5, 9·5)NS
Total phosphorus (p.p.m.)NS 98·4* (93·2, 103·6)
Available phosphorus (p.p.m.)NS  5·7* (5·3, 6·2)
Calcium (meq 100 g−1)  0·8 (0·6, 1·0)  0·4 (0·3, 0·5)
Magnesium (meq 100 g−1)  0·2 (0·2, 0·2)  0·1 (0·1, 0·1)
Potassium (meq 100 g−1)  0·0 (0·0, 0·0)  0·0* (0·0, 0·0)
Sodium (meq 100 g−1)  0·1 (0·1, 0·1)  0·1 (0·1, 0·1)

Combined species seedling survival in experiment 2 averaged 61%, 20 months after planting (CL = 55·0–66·1). Mulch application significantly enhanced overall seedling survival, with surviving seedlings healthier in mulched treatments (Table 4). Survival of planted A. l. sophorae seedlings was enhanced 22% by mulch application (Table 5). Mulching alone doubled the survival of B. i. integrifolia, although this effect was somewhat reduced by herbicide application (Table 5). Of the remaining B. i. integrifolia, there was no significant effect of mulch or herbicide on individual seedling health. Planting density only affected the survival of C. glauca, with survival being reduced from 89% and 90% in the 1 seedling m−2 treatment to 73% with a planting density of 2 seedlings m−2 (F = 11·8, d.f. = 2, 22, P < 0·001), while health of surviving seedlings was more than doubled with mulch (Table 5). Although only 16% of A. smithii seedlings were alive after 20 months, survival and health were enhanced in mulched treatments (Table 5). Mulch did not affect the growth of B. i. integrifolia (Table 6) but significant mulch effects were evident in changes to basal girth, canopy width and height of A. l. sophorae and C. glauca seedlings after 20 months. Mulch application to seedlings of A. l. sophorae and C. glauca enhanced basal girth by 64% and 49%, respectively, canopy width by 77% and 166%, respectively, and height by 29% and 60%, respectively (Table 6).

Table 4.  Planted seedling survival and health, pooled across species, in the second experiment in December 2002
Plot treatmentAlive (%)Healthy (%)
  1. Statistics refer to the planned contrast, mulch vs. non-mulch, after a main effect was detected. Predicted means are presented with upper and lower 95% confidence limits in parentheses. Means in the same column with a different superscript differ significantly.

Mulch + herbicide68·4a (57·6, 77·5)79·7 (62·2, 90·4)
Mulch75·3a (65·3, 83·1)86·8 (72·8, 94·1)
Non-mulch + herbicide48·4b (37·2, 59·8)74·3 (53·8, 87·7)
Non-mulch52·2b (36·8, 59·1)57·7 (36·2, 76·7)
StatisticsF = 32·2, d.f. = 1, 8, P= 0·005F = 4·8, d.f. = 1, 8, P= 0·059
Table 5.  Species-specific planted seedling survival and health after 20 months in the second experiment. Statistics refer to the planned contrasts after a main effect was detected
TreatmentA. l. sophoraeB. i. integrifoliaC. glaucaA. smithii
A (%)H (%)A (%)H (%)A (%)H (%)A (%)H (%)
  • Predicted means are presented with upper and lower 95% confidence limits in parentheses unless there was 100% survival within a group or statistical analyses were not possible (NA), in which case observed means ± 1 SE are presented.

  • NS, non-significant. Superscripts identify statistically different means within a column.

  • A, alive; H, healthy.

  • *

    The C. glauca survival models only converged when averaged across mulching and non-mulching treatments.

Mulch + herbicide 91·2a (80·7, 96·3) 88·1 (77·1, 94·3)66·4b (49·1, 80·1)68·7 ± 8·0*84·6a (59·4, 95·4)24·3a (9·1, 50·9)39·2 ± 10·9
Mulch100a10085·3a (71·2, 93·2)88·8 ± 6·292·1 (78·5, 97·4)87·8a (65·2, 96·5)30·3a (12·4, 57·3)40·7 ± 12·1
Non-mulch + herbicide 71·4b (55·3, 83·4)10030·1c (16·9, 47·7)79·1 ± 14·7*38·2b (14·3, 69·8) 7·1b (1·7, 24·9)17·4 ± 8·9
Non-mulch 84·9b (70·9, 92·9) 79·2 (64·9, 88·8)31·1c (17·9, 48·3)59·2 ± 15·974·5 (50·7, 89·2)32·5b (11·8, 63·3) 0·3b (0·0, 0·3) 5·6 ± 5·6
StatisticsF = 9·2, d.f. = 1, 8, P= 0·02NSa v b: F= 6·6, d.f. = 1, 7, P= 0·04 ab v c: F= 41·0, d.f. = 1, 7, P < 0·001NAF = 5·0, d.f. = 1, 8, P= 0·06F = 18·8, d.f. = 1, 8, P= 0·003F = 13·6, d.f. = 1, 8, P= 0·006NA
Table 6.  Growth of planted seedlings in the second experiment after 20 months
TreatmentA. l. sophoraeB. i. integrifoliaC. glauca
BG (mm)CW (cm)H (cm)BG (mm)CW (cm)H (cm)BG (mm)CW (cm)H (cm)
  1. Statistic refers to the planned mulch vs. non-mulch contrast after a main effect was detected. Predicted means are presented with upper and lower 95% confidence limits in parentheses. Superscripts identify statistically different means within a column.

  2. BG, basal girth; CW, canopy width; H, height.

  3. NS, not significant.

Mulch + herbicide23·0a (18·4, 28·9)190·9a (145·7, 250·1)141·5a (126·6, 158·2)8·1 (9·8, 9·6)28·8 (23·8, 34·9)70·7 (61·6, 80·4)8·5a (7·4, 9·8) 42·4a (34·5, 52·1)111·3a (101·0, 122·2)
Mulch20·9a (9·5, 15·2)165·5a (125·7, 217·9)131·0a (116·5, 147·4)7·1 (6·0, 8·2)23·7 (20·1, 27·9)66·0 (58·5, 74·1)9·5a (8·3, 10·9)46·8a (38·4, 57·1)112·1a (102·0, 122·7)
Non-mulch + herbicide12·0b (9·5, 15·2) 87·6b (65·9, 116·3) 92·4b (81·5, 104·8)9·1 (7·1, 11·4)26·9 (20·3, 35·6)74·3 (61·1, 88·9)6·1b (5·3, 7·1)17·9b (14·3, 22·5) 71·9b (62·8, 81·7)
Non-mulch14·8b (11·8, 18·7)114·0b (86·4, 150·3)118·2b (105·0, 133·2)8·6 (6·7, 10·8)24·1 (18·3, 31·6)79·3 (66·1, 93·7)6·0b (5·2, 6·9)15·6b (12·6, 19·4) 67·9b (59·5, 76·8)
StatisticF = 16·9, d.f. = 1, 8, P= 0·003F = 14·7, d.f. = 1, 8, P= 0·005F = 8·4, d.f. = 1, 8, P= 0·02NSNSNSF = 37·0, d.f. = 1, 8, P < 0·001F = 84·7, d.f. = 1, 8, P < 0·001F = 74·0, d.f. = 1, 8, P < 0·001

Native herbaceous cover was dominated by I. cylindrica, and was affected by the establishment treatments (F = 3·0, d.f. = 5, 40, P = 0·02). Planting seedlings reduced native herbaceous cover by 23% (F = 7·5, d.f. = 1, 48, P = 0·009); this was largely an effect of halving I. cylindrica cover in those subplots where seedlings were planted at 1 m−2 compared with unplanted subplots (F = 19·3, d.f. = 1, 48, P < 0·001; Fig. 1). Recruitment of native woody species that were not experimentally provided, in the form of seed or seedlings, was negligible across the sampling period. Seeding with A. l. sophorae affected the regenerating cover of native woody species differently in mulched and non-mulched plots (plot : subplot interaction: F = 6·1, d.f. = 15, 40, P < 0·001; Fig. 2), seeding into mulch being the effective treatment. Total plant cover differed across subplot treatments (F = 7·8, d.f. = 5, 40, P < 0·001). Planting native woody seedlings at 1 m−2 markedly increased total plant cover 20 months after planting, from 22% in unplanted subplots (CL 8·4–62·4; establishment treatments ii and vi, respectively) to 37% in planted subplots (CL 12·8–94·2; establishment treatments i and iii, respectively; F= 20·3, d.f. = 1, 48, P < 0·001). This was because of the presence and growth of planted seedlings and their contribution to total plant cover in the planted subplots.

Figure 1.

Effect of planted seedlings on regeneration of I. cylindrica after 20 months in the second experiment. Predicted means with 95% confidence intervals are presented. Pairs of bars with different superscripts identify a significant planned contrast.

Figure 2.

Native woody regeneration cover from seeding of A. l. sophorae in mulched plots after 20 months in the second experiment. Predicted means with 95% confidence intervals are presented.

As weed control only affected one of the response variables reported thus far, weed control was removed from the model, pooling M into mulched and non-mulched plots only. Planting seedlings significantly enhanced subcanopy cover development (F = 6·84, d.f. = 1, 54, P = 0·01) but only in mulched plots (plot : subplot interaction: F= 4·2, d.f. = 5, 50, P = 0·003; Fig. 3). The maximum subcanopy cover that developed in mulched and high-density planted subplots was 45%. Low-density planted subplots only achieved a subcanopy cover of 26% in mulched plots, although this did not differ significantly from high-density plantings (Fig. 3).

Figure 3.

Effect of mulch application and establishment treatment on subcanopy cover development after 20 months in the second experiment. Predicted means and 95% confidence intervals are presented.


Survival of wet sclerophyll and rainforest seedlings averaged 26% after 28 months in the first experiment. Mixed plantings of mature and late secondary rainforest species on the NSW north coast have achieved survival rates of 80–95% after 4 years and have displayed rapid height and diameter growth (Kooyman 1996). After 30 months, survival of eucalypt and Acacia seedlings in an Indonesian I. cylindrica grassland averaged 91% (Turvey 1996), highlighting the limited success of seedling establishment in the first experiment of this study.

Soil chemical properties at the I. cylindrica sites were impoverished and similar to those of a quarry in Hong Kong where rehabilitation plantings also failed (Jim 2001). Organic matter at the BBNP sites was less than half of that in the quarry sites, where meagre organic matter was related to the site's inability to support vegetation. Soil nitrate levels at the BBNP sites were approximately 10% of the minimum levels generally recommended for plant growth (Cumming & Elliott 1991). Organic matter content and nutrient pools in the I. cylindrica sites were, on average, only 25% of undisturbed littoral rainforest and E. grandis rainforest ecosystems on the Bongil Peninsula sand mass (Cummings 2003). This degree of abiotic limitation generally restricts restoration plantings and natural regeneration (Whisenant 1999). Successful establishment of seedlings can be retarded further by browsing, which limits a seedling's capability to accumulate nutrient and energy reserves (Whisenant 1999; Sweeney, Czapka & Yerkes 2002). Although it was not measured, browsing of unprotected seedlings by wallabies (W. bicolor) deleteriously affected seedling growth and survival in this study. Drought-like conditions probably exacerbated the combined effects of wallaby grazing and soil abiotic limitations on seedling survival and establishment.

Initial burning and weed control reduced the cover of both native and introduced woody species. Limited regeneration of both native and introduced woody species meant no effect of weed control being detected over the 2-year period. Burning and control of woody weeds did not affect the regeneration of native herbaceous cover, although a slight increase in I. cylindrica cover was recorded over time. Although burning favours the spread of I. cylindrica (King & Grace 2000), establishment of woody vegetation reduces I. cylindrica cover (Otsamo et al. 1995; Dozier et al. 1998). Establishment failure of planted seedlings after burning allowed the re-establishment of I. cylindrica. It was subsequently hypothesized that an abiotic threshold (sensu Friedel 1991; Laycock 1991) related to soil limitations restricted establishment of native planted and regenerating species (Whisenant 1999) and therefore restoration of the I. cylindrica area in BBNP. Biotic interactions (wallaby browsing and blady grass competition) are likely to have further retarded establishment of planted native woody species. The second experiment tested this hypothesis by addressing the soil abiotic limitations, while eliminating wallaby browsing and partially suppressing blady grass.

Immediate impacts of mulch on soil chemical parameters were difficult to detect in the second experiment. While mulching did improve soil potassium and pH, anticipated increases in organic matter and soil moisture were not detected. After treatment with an organic amendment, increases in pH, water retention and cation exchange capacity were detected in the reclamation of a Canadian sand mine (Fierro, Angers & Beauchamp 1999). Improvements in soil structure, stability, moisture capture and retention, fertility and biological activity are usually achieved with organic amendments (Bradshaw 1983; Whisenant 1999). Possible reasons for failing to detect significant soil chemical improvements include rapid leaching from sands, rapid uptake by plants and/or the sampling strategy employed.

Impacts of mulch on planted seedlings survival were readily detected, through enhanced survival, health and growth of each of the sclerophyllous species. Organic amendments have rarely been used to rehabilitate Imperata grasslands. However, application of biosolids to sand-mined soils improved plant biomass in western Australia (Rate, Lee & French 2004). In derelict mine tailings and waste rock treated with dolomite, 2 cm of surface mulch improved survival, growth and cover of 12 native Australian species (Grant, Campbell & Charnock 2002). Mulching reduced weed cover and increased growth of two eucalypt species on the NSW Northern Tablelands (George & Brennan 2002). In saline-affected areas, mulch improved the survivorship and growth of a suite of native sclerophyllous species, including Acacia stenophylla A. Cunn. ex Benth. and Casuarina cunninghamiana Miq. ssp. cunninghamiana (Marcar et al. 2000). While the mechanism by which mulch application enhanced growth of desirable native species remains unclear, the positive benefits are obvious.

Irrespective of mulching, use of planted seedlings halved the regenerating cover of I. cylindrica. The establishment of desirable competitive species is important in controlling and replacing I. cylindrica where it dominates degraded sites (Dozier et al. 1998; Ramsey et al. 2003). Acacia mangium Willd. plantations were used to enhance indigenous tree and shrub species establishment in Indonesian I. cylindrica grasslands (Otsamo 2000). Establishment of the perennial shrub Artemisia californica in organic amended plots allowed it to out-compete exotic annuals for nutrients and water (Zink & Allen 1998). Integrated management of I. cylindrica using a suite of control techniques has been advocated (Dozier et al. 1998). Within the first 12 months of the second experiment, mulch reduced the cover of regenerating native herbaceous species (Vile 2002). After 20 months this effect was no longer apparent, with I. cylindrica (and all native herbaceous) cover similar across mulched and non-mulched treatments. Mulch may limit competition between planted seedlings and I. cylindrica during the establishment phase. Therefore, the benefits of mulch for seedling establishment may be twofold: (i) enhancing soil limitations related to fertility and structure; and (ii) initially restricting regeneration of, and competition from, I. cylindrica.

Differences in the best-case survival of planted seedlings of A. l. sophorae (91%), C. glauca (92%) and B. integrifolia (85%) compared with A. smithii (30%) highlight the importance of selecting species appropriate to the site for restoration. Plot and subplot treatments also had different effects on the planted and seeded species. Seeding of A. l. sophorae was only successful in mulched plots, planting density affected C. glauca survival, and post-planting herbicide application reduced B. i. integrifolia survival. Elsewhere, seedling establishment from broadcast seed was enhanced with mulch application (Rockich et al. 2002), a population of an endangered shrub was affected by Glyphosate spray drift (Matarczyk et al. 2002), and planting density effects on plantation trees are commonly observed (Neilsen & Gerrand 1999). Clearly, species were affected differently by different treatments in this study. By conducting research trials using different species and management techniques, managers can make informed decisions about how their actions may influence community composition.

management and theoretical implications

Native woody canopy development was maximized in this trial by planting native seedlings within mulched plots. By identifying abiotic and biotic conditions under which native species can colonize, ecological restoration processes can be enhanced (Elmarsdottir, Aradottir & Trlica 2003). Establishment of a native woody canopy cover is crucial in not only controlling I. cylindrica (Dozier et al. 1998) but also in enhancing ecosystem function. Enhancing woody plant cover can reduce wind erosion, improve energy capture, increase litter fall, increase soil stability and improve soil moisture retention and microclimatic conditions for germination of other desirable species (Whisenant 1999). Therefore, by creating conditions that enhance survival of desirable native species in the short term by identifying and addressing establishment barriers, not only can ecosystem function be repaired, successional processes that may lead to longer term goal communities can also be initiated.

State-and-transition approaches to ecological restoration have been advocated (Yates & Hobbs 1997). Degradation across a restoration threshold can occur with far less effort than that required to reverse the degradation (Hobbs & Norton 1996). Restoration thresholds have been divided into those that represent a transition because of abiotic limitations and those because of biotic interactions (Whisenant 1999). Changes between states can be measured by changes in variables reflecting ecosystem primary processes, composition and structure (sensu Aronson et al. 1993; Whisenant 1999). Assessment of ecosystem function variables (Cummings 2003) and soil chemical response variables (this study) indicated that an abiotic transition threshold may have restricted restoration efforts in the first experiment. Wallaby browsing and increasing competition from I. cylindrica were probable biotic interactions that further restricted restoration.

Treatment of both abiotic limitations and biotic interactions simultaneously resulted in restoration success, with improved establishment and growth of desirable species. These results support the notion that a transition threshold exits in the unmanaged BBNP I. cylindrica grasslands, consisting of both abiotic and biotic components. A state-and-transition model incorporating the findings of this study and other research in nearby areas has been developed for restoring biodiversity to degraded areas in BBNP (Cummings 2004). Application of this model will result in enhanced restoration of these degraded areas in the conservation estate, adding integrity to the reserve system by enhancing native woody cover and restricting recruitment and establishment of weed species.

The second experiment is the most recent attempt at restoration of this disturbed site. At least four separate previous attempts to restore these sites had occurred and failed (the first experiment and three earlier attempts; M. Smith, personal communication). Accurately identifying and managing barriers to restoration is essential to allow their management and to enhance the likelihood of restoration success (Hobbs & Norton 1996). Our experiments identified the need for both soil and vegetation management at these degraded sites to initiate the restoration process. By implementing an adaptive management approach, information gained from unsuccessful restoration attempts and experimental trials was used to determine a successful restoration strategy for the site. Management recommendations from this research will be incorporated into the reserve plan of management. Further, the success of this adaptive approach has highlighted to the land managers the usefulness of logical and sequential experiments in guiding management decisions.


Martin Smith, Glenn Storrie and Alan Jeffery from the NSW NPWS are acknowledged for their continued support of this research. Mark Vile is acknowledged for collecting and analysing soil samples from the second experiment. NPWS field crews are thanked for their efforts in establishing the trials. The research was funded by the NSW NPWS and the Australian Research Council; J. Cummings received a DEETYA-funded scholarship for this research. Three referees are thanked for their comments, which improved the manuscript.