Restoring ecological function in temperate grassy woodlands: manipulating soil nutrients, exotic annuals and native perennial grasses through carbon supplements and spring burns

Authors

  • SUZANNE M. PROBER,

    Corresponding author
    1. Institute for Land, Water and Society, Charles Sturt University, PO Box 789, Albury, NSW 2640, Australia; and
      Suzanne M. Prober, Institute for Land, Water and Society, Charles Sturt University, PO Box 789, Albury, NSW 2640, Australia (e-mail suzanne.prober@bigpond.com).
    Search for more papers by this author
  • KEVIN R. THIELE,

    1. Institute for Land, Water and Society, Charles Sturt University, PO Box 789, Albury, NSW 2640, Australia; and
    Search for more papers by this author
  • IAN D. LUNT,

    1. Institute for Land, Water and Society, Charles Sturt University, PO Box 789, Albury, NSW 2640, Australia; and
    Search for more papers by this author
  • T. B. KOEN

    1. Centre for Natural Resources, Department of Infrastructure, Planning and Natural Resources, Cowra, NSW 2794, Australia
    Search for more papers by this author

Suzanne M. Prober, Institute for Land, Water and Society, Charles Sturt University, PO Box 789, Albury, NSW 2640, Australia (e-mail suzanne.prober@bigpond.com).

Summary

  • 1Ecological invasions are often associated with persistent changes to underlying ecological processes. Restoration of invaded communities is dependent on manipulation of these processes to favour the target species composition and impart resistance to further invasion. We applied these principles to extensively degraded grassy woodlands in temperate agricultural regions of Australia, where widespread invasion by mediterranean annuals is related to altered ecological processes such as soil nutrient cycling.
  • 2We investigated carbon supplementation and spring burns, in association with re-establishment of native perennial grasses, as potential management tools for manipulating nitrogen cycling, soil seed banks and establishment conditions in degraded woodland understoreys. Through these mechanisms we aimed to enhance native cover and increase resistance to invasion by exotic annuals.
  • 3In two contrasting degraded remnants, repeated sucrose applications temporarily reduced soil nitrate to inferred pre-European levels, which dramatically reduced growth of exotic annuals and enhanced native perennial abundance. Repeated spring burns did not reduce soil nitrate but reduced exotic annual grasses through effects on soil seed banks and/or establishment conditions.
  • 4Spring burns and carbon supplements both significantly enhanced establishment of Themeda australis, a dominant tussock grass prior to European settlement. Combinations of T. australis seed addition and either spring burning or carbon supplements enhanced native plant abundance more effectively than treatments without seed addition.
  • 5Within 18 months of their establishment, T. australis seedlings significantly reduced soil nitrate in some treatments, providing a preliminary indication that re-establishment of a dense sward of this species may restore ecosystem function to a low-nitrate state that favours native perennials over exotic annuals.
  • 6Synthesis and applications. Ecological restoration can be viewed as targeted intervention in species–environment interactions, whereby ecological conditions are manipulated to enhance establishment or vigour of key species, and these species in turn help restore ecological processes that favour the target species composition. In grassy ecosystems re-establishing a perennial sward of appropriate native tussock grasses may be critical for restoring pre-disturbance nitrogen cycles and improving resistance to invasion by exotic annuals. Carbon supplements and spring burns facilitate this process through complementary mechanisms.

Introduction

Ecological restoration for biodiversity conservation generally seeks to shift species composition towards a state more closely resembling a reference ecological community, often the historic indigenous community (Aronsen et al. 1993; Hobbs & Norton 1996). Re-establishing a desired species composition is rarely straightforward and often depends on re-instating interacting ecological processes that have become disrupted or altered during previous degradation phases. Consequently, determining effective management interventions that restore ecosystem function is often essential for achieving restoration goals.

The need to restore ecosystem function is especially relevant to temperate grassy woodlands of southern Australia. During nearly 200 years of European settlement, these woodlands have been extensively fragmented and degraded through clearing, livestock grazing and cultivation. This has led to widespread changes in their condition and species composition that are closely linked with changes in ecological processes such as soil nutrient cycling (Prober, Thiele & Lunt 2002), hydrology (Hobbs 1993), biological invasions (Moore 1953; Hobbs 1989), plant–animal interactions (Landsberg, Morse & Khanna 1990) and tree regeneration (Yates & Hobbs 1997).

Changes to woodland understoreys are of particular concern. Across the winter rainfall areas that support these woodlands, remnants have been invaded by exotic annuals. The naturally diverse understorey of perennial grasses and forbs has become depauperate, and many once common native plant species have become localized and isolated (Prober & Thiele 1995; Hobbs & Yates 1999). The widespread loss of deep-rooted perennials has contributed to broad-scale landscape problems such as salinity, soil erosion and tree decline (Landsberg, Morse & Khanna 1990; Hobbs 1993).

At the site scale, the change from perennial natives to exotic annuals has modified internal ecological processes in many degraded grassy woodlands. Soil seed banks have become dominated by cool-season exotic annuals (Lunt 1990) that establish dense swards each autumn–winter and limit establishment of native perennials (Lenz 2004). One potential method for breaking the cycle of seed bank replenishment and seedling establishment in fire-intolerant annuals is to burn in spring before the annuals mature (Menke 1992; Prober, Thiele & Koen 2004). This may subsequently promote native perennials that are naturally adapted to occasional burning (Lunt & Morgan 2002).

Dominance by annuals is also associated with a shift in internal soil nitrogen cycling, involving seasonally enhanced soil nitrate levels that favour the establishment of annuals over native perennials (Prober, Thiele & Lunt 2002). Increased concentrations of available soil nitrogen are commonly associated with early seral species during secondary succession (Parrish & Bazzaz 1982; McLendon & Redente 1992). However, in Australian grassy woodlands exotic annuals often persist in the longer term, even after exogenous disturbances such as livestock grazing have ceased (Prober, Thiele & Lunt 2002; Spooner, Lunt & Robinson 2002).

Restoring pre-disturbance nitrogen cycles may thus be essential for successful grassland restoration. Techniques for reducing soil-available nitrogen are not well established but include repeated burning (Ojima et al. 1994) and addition of carbon to the soil. Carbon supplements enhance the activity of carbon-limited decomposing soil micro-organisms that subsequently compete with plants for available soil nitrogen (Jonasson et al. 1996). Carbon supplements have been shown to immobilize soil nitrogen and limit growth of nitrophilic plant species (particularly annuals or early successional species) in studies in the northern hemisphere (Reever Morghan & Seastedt 1999; Török et al. 2000; Blumenthal, Jordan & Ruselle 2003) but the technique has been little used in Australia. Unfortunately, native species have not always been enhanced by carbon supplements (Reever Morghan & Seastedt 1999; Alpert & Maron 2000) and repair of ecosystem function has rarely been demonstrated (McLendon & Redente 1992; Bakker & Wilson 2004).

We investigated carbon (sucrose) supplementation and spring burning, in association with re-establishment of native perennial grasses, as potential management interventions for restoring ecological function and composition in degraded grassy white box Eucalyptus albens woodlands in temperate south-eastern Australia. Our specific objectives were to (i) determine the influence of management treatments on soil nutrients and exotics; (ii) assess associated effects on abundance and establishment of native perennials; and (iii) evaluate subsequent effects of modified vegetation on soil nutrients, and hence the potential effectiveness of the management interventions for imparting long-term resistance to invasion by exotic annuals.

Methods

experimental design

Restoration trials were established at two sites (Windermere and Green Gully) in central New South Wales, selected to represent two common degradation states in grassy white box woodlands (see below; Prober, Thiele & Lunt 2002). The soils, topography and climate of these sites were typical of remnant woodlands in the region, as determined from extensive floristic and soil surveys (Prober & Thiele 1995; Prober, Thiele & Lunt 2002). The sites were located in undulating terrain on red duplex soils (moderately acidic sandy loams to sandy clay loams over clay) derived from granodiorites (Prober, Lunt & Thiele 2002). Average annual rainfall is 650 mm, although this is highly variable, with drought periods every 10–20 years. Experimental plots were placed in historically cleared areas, to avoid the complicating influence of trees on soil nutrients and floristic composition (Prober, Lunt & Thiele 2002), and were fenced to exclude livestock.

The Windermere site (34°11′S, 148°33′E) was on an upper slope within grazed native pasture. Initial understorey condition represented the Bothriochloa macra degradation state (Prober, Thiele & Lunt 2002), dominated by Bothriochloa macra and other native perennial grasses that increase under livestock grazing (Moore 1953). Cool-season exotic annuals, particularly grasses of the genera Vulpia and Bromus and a range of broad-leaf forbs, predominated between the grass tussocks, with few native forbs or perennial exotics.

The Green Gully site (34°18′S, 148°03′E) was on a lower slope within an intermittently grazed travelling stock reserve. Initial understorey condition represented the exotic annual degradation state (Prober, Thiele & Lunt 2002), dominated by robust cool-season exotic annual grasses, particularly Avena barbata and Bromus diandrus, with occasional broad-leaf annuals (mainly Echium plantagineum), and very sparse native perennials.

At Windermere, a three plot treatment × three seed treatment × four replicate factorial trial was established using 3 × 2-m plots separated by 1-m buffers arranged in a randomized complete block design. Plot treatments included an untreated control, annual spring burns and regular carbon supplements. Thorough burns (that removed much of the litter layer and exposed bare soil) were achieved using a gas-powered weed burner in mid-October 2002 and 2003, when annuals had high biomass but before they had set seed. For carbon (C) supplements, sucrose (white sugar) was applied to the soil surface at 0·5 kg m−2 every 3 months (8·4 t C ha−1 annum−1), beginning in May 2002.

Seed treatments included an unseeded control and surface sowing with the summer-active C4 native perennial grass Themeda australis (hereafter Themeda) or the cool-season C3 native perennial grass Poa sieberiana (hereafter Poa). These two species dominated woodland understoreys before European settlement (Prober & Thiele 1995) but were absent from both sites at the start of the trial. For Themeda, 250 kg ha−1 seed and chaff (approximately 4000 seeds plot−1) were applied in August 2002 and 2003. For Poa, 4 kg ha−1 seed and chaff (approximately 8000 seeds plot−1) were applied after burning in October 2002 and 2003 and in May 2003.

An identical trial was established at Green Gully except for the addition of a herbicide treatment, leading to a four plot treatment × three seed treatment × four replicate factorial trial. For the herbicide treatment, we killed all top growth using the non-selective contact herbicide BASTA™ (active component glufosinate ammonium, Bayer CropScience Pty. Ltd., East Hawthorn, Vic., Australia) at 1 L ha−1 in early October 2002 and 2003. This treatment was intended to mimic the effect of burning on exotic seed pools without influencing organic matter, soil nutrients or establishment conditions, to help identify which processes influenced the outcomes of burning. The herbicide was also applied to burnt plots at Green Gully, to make burning easier.

floristic monitoring

Floristic composition was monitored in November 2001 (prior to any treatments), October 2003 (before the second burn) and autumn 2004 using a point-intercept technique (modified from Everson & Clarke 1987). An 8-mm dowel was placed vertically at each of 50 points on a grid across each plot; the relative abundance (points) for any species was the number of points at which any of its leaves, stems or inflorescences intercepted the dowel. Species that were present but did not intercept the dowel at any point were allocated a nominal abundance of 0·5. Nomenclature follows Harden (1990–93) and Wheeler, Jacobs & Whalley (2002).

Plant biomass was measured in October 2003 by harvesting three subsamples (totalling 0·1 m2) per plot. Plants were cut at ground level and sorted into relevant floristic components, then weighed after oven drying at 60 °C for 24 h. Density of individuals in each floristic group was estimated by counting all harvested plants. Average plant size was derived from density and biomass data. Composite estimates of the biomass and cumulative abundance of various species groups were also calculated. The density of sown native grasses was estimated at occasional intervals by counting plants within four to six 0·25-m2 subplots in each plot.

leaf nitrogen and carbon

Leaf nitrogen and carbon percentages were measured for Bromus diandrus (a dominant exotic annual grass) on control and carbon-supplemented plots at Green Gully in September 2002. The third newest leaf was harvested from each of 50 individuals within each plot. Samples were oven dried at 85 °C for 16 h, then ground to 0·1 mm and analysed for nitrogen and carbon using a modified Pregl-Dumas method.

soil analyses

Six 2-cm diameter soil cores to 10 cm depth were collected from random points (at least 2 cm from any perennial tussock) within each plot every 3 months. Soil samples were stored on ice, then sent to commercial laboratories for analysis. Samples from each plot were mixed, dried at 40 °C and ground to pass through a 2-mm sieve. Analyses were undertaken on each bulked sample as follows (method numbers apply to Rayment & Higginson 1992). Nitrate and ammonium (2 m KCl, 7C2 but with cadmium column reduction) were measured every 3 months, either for all plots (May 2002 prior to treatment, November 2002, February 2003, May 2003, April 2004) or subsets of plots (control and carbon-supplemented plots only, August 2002; unseeded plots only, August 2003, November 2003). Available phosphorus (Olsen's bicarbonate extractable, 9C2) was measured prior to treatment (May 2002) and in November 2002, May 2003 and April 2004. Total nitrogen (modified Dumas method; LECO Corporation 1995) and total carbon (induction furnace, 6B3, without correction for carbonates; visual carbonates were negligible) were measured at the final measurement date, April 2004.

data analysis

Data for each measurement date and site were analysed using GenStat (GenStat 2003) by analysis of variance of the factorial, randomized complete block designs (or relevant partial designs). Where appropriate, raw data were transformed using the natural log(ln(x + 1)) or logit(ln(x)/ln(1 − x)) transformations, and pre-treatment data were used as covariates to correct for initial variation. Treatment means were compared using Fisher's protected least significant differences (Steel & Torrie 1981). Back-transformed means are presented.

Results

Severe drought during much of the study period led to failed Poa establishment and reduced survival and growth of Themeda. Poa treatments could thus be considered identical to unseeded treatments. Also, herbicide application did not eliminate seeding of annuals because of early maturation in drought conditions.

effects of plot treatments on soil nutrients

Soil nitrate fluctuated seasonally under most treatments, with distinct peaks in summer–autumn (Fig. 1). This pattern reflected the growth of cool-season annuals, with high nitrogen use in winter and spring as plants establish and mature, and re-incorporation of nitrogen into the soil during summer after plants have senesced (Jones & Woodmansee 1979; Jackson et al. 1988; Prober, Thiele & Lunt 2002).

Figure 1.

Plot treatment effects on soil nitrate across the experimental period. Significance levels (*P < 0·05; **P < 0·01; ***P < 0·001; not shown, not significant) refer to comparisons within measurement dates; different letters indicate significant differences between plot treatment means within measurement dates (P < 0·05). Note that the ‘burn’ treatment at Green Gully was preceded by a herbicide application and that only subsets of plots were measured for August 2002, August 2003 and December 2003 (see the Methods). See Fig. 6 for seed treatment effects and interactions.

Carbon supplements dramatically reduced soil nitrate, especially in summer–autumn when peaks of up to 30 mg kg−1 were suppressed to < 2·5 mg kg−1 (Fig. 1). One exception occurred at Green Gully (Fig. 1b), where extremely low rainfall after the March 2004 carbon addition may have limited incorporation of sugar into the soil. Burning led to slight (Windermere) and moderate (Green Gully) increases in soil nitrate relative to controls in the first season, but no nitrate flushes could be attributed to the second burn (Fig. 1).

Treatments had relatively weak and erratic effects on other soil nutrients. Carbon supplements frequently but inconsistently suppressed ammonium (by up to 6 mg kg−1 compared with controls; Fig. 2), occasionally reduced phosphorus by small amounts (1–2 mg kg−1; Fig. 3), significantly increased total carbon and marginally increased total nitrogen by the end of the trial (Table 1). Burning increased ammonium at Green Gully in the first season but reduced ammonium between June and December 2003 at Windermere (Fig. 2). Burning did not significantly influence phosphorus, total carbon or total nitrogen, although total carbon and nitrogen were lower on burnt than control plots by the end of the trial (Fig. 3 and Table 1). Herbicide application had minimal effects on soil nutrients, except for a brief decline in phosphorus after the first herbicide application (Fig. 3b) and a small increase in nitrate on one measurement date (Fig. 1b).

Figure 2.

Plot treatment effects on soil ammonium across the experimental period. Significance levels (*P < 0·05; **P < 0·01; ***P < 0·001; not shown, not significant) refer to comparisons within measurement dates; different letters indicate significant differences between values within measurement dates (P < 0·05). Note that only subsets of plots were measured for August 2002, August 2003 and December 2003.

Figure 3.

Plot treatment effects on soil phosphorus across the experimental period. Significance levels (*P < 0·05; **P < 0·01; ***P < 0·001; not shown, not significant) refer to comparisons within measurement dates for effects of plot treatment; different letters indicate significant differences between values within measurement dates (P < 0·05).

Table 1.  Plot treatment effects on soil (April 2004) and leaf (Bromus diandrus; August 2002) carbon and nitrogen. Different letters indicate significant differences between means (P < 0·05); NS, not significant
 Site P Spring burnUntreated controlCarbon additionHerbicide
Soil total C (%)Windermere< 0·0011·358a 1·475a 1·850b
Green Gully    0·0061·925a 2·158a 2·508b2·108a
Soil total N (%)Windermere    0·0110·122a 0·130ab 0·143b
Green Gully    0·0970·170a 0·190a 0·202a0·183a
Leaf C%Green GullyNS45·33a45·30a
Leaf N%Green Gully< 0·001 1·925a 1·467b

effects of plot treatments on floristic composition

Carbon supplements strongly suppressed all exotic annual species and species groups, leading to highly significant reductions in total exotic biomass and relative abundance (Fig. 4 and Table 2). These effects were because of reduced growth of individuals (P < 0·001) rather than a decrease in the number of individuals establishing. Foliage analysis for Bromus diandrus indicated a significant decline in nitrogen (Table 1) but no change in total carbon as a result of carbon supplements.

Figure 4.

Effects of plot treatments on relevant biomass classes (spring 2003) and plot × seed treatments on relative native perennial grass abundance (points, autumn 2004). Different letters indicate significant differences (P < 0·05) between values within biomass classes. For native perennial grass abundance, asterisks (*P < 0·05, ***P < 0·001) indicate significance of plot treatment (T), seed treatment (S) and site–seed treatment interaction (T×S); least significant differences for perennial grass abundance (for T×S, P < 0·05) are indicated by error bars.

Table 2.  Plot treatment effects on relative abundance of plant species and groups for spring 2003 and autumn 2004. Data for Digitaria hubbardii could not be analysed because of high numbers of zeros in some treatments. Different letters indicate significant differences between means (P < 0·05); NS, not significant
 YearBurn (points)Control (points)Sucrose (points)Herbicide (points) P
  • *

    For Themeda plots only.

  • Treatment–seed interaction significant, see text.

Windermere
Native grasses
Austrodanthonia spp.2003 0·6a  2·3b 1·9b     0·008
2004 0·5a  2·1b 2·3b     0·003
Bothriochloa macra 2003 6·0a  7·9a16·9b     0·001
200425·8a 27·5ab32·2b     0·037
Digitaria hubbardii 2004 7·2  4·0 1·3 NA
Elymus scaber 2003 5·9a 11·8b 5·4a     0·007
2004 1·0  2·1 0·9 NS
Themeda australis * 2003 2·2  0·6 2·3 NS
200411·2a  3·3b11·6a     0·027
Total native grasses200318·7a 28·8b31·3b < 0·001
200442·7a 38·8b44·2a     0·015
Exotic annuals
Erodium botrys 200319·5a 15·1a 2·5b < 0·001
Hypochaeris glabra 200310·5a  8·8a 4·4b     0·003
Trifolium subterraneum 200327·0a 18·3b 8·7c < 0·001
Legumes200332·2a 21·3b10·2c < 0·001
Annual grasses2003 2·4b  7·7a 0·5c < 0·001
Broad-leaf annuals200376·1a 61·3b21·4c < 0·001
Total exotic annuals200380·1a 70·3b22·5c < 0·001
Green Gully
Native grasses
Themeda australis * 200428·5a  5·9b23·8a  1·3c< 0·001
Total native grasses2003 0·8ab  0·3a 1·0b  0·4a    0·025
Total native grasses200416·1a  5·8b19·6a  5·7b< 0·001
Exotic annuals
Avena barbata 200329·0a 48·7b27·5a 46·3b< 0·001
Bromus diandrus 2003 2·1a 34·4c11·4b 35·3c< 0·001
Echium plantagineum 200346·7a  8·8bc 5·5c 15·4b< 0·001
Annual grasses200333·5a 87·5c42·1b 84·3c< 0·001
Broad-leaf annuals200348·8a 13·9bc11·8c 18·0b< 0·001
Total exotic annuals200382·1b100·8a54·7c102·7a< 0·001

Reduced exotic growth on carbon-supplemented plots led to significant increases in native perennial grass abundance by autumn 2004. At Green Gully, the average increase was considerable (237%; Table 2) and was particularly great on plots seeded with Themeda (414%, interaction P < 0·001; Fig. 4b). In seeded plots, most of the increase in native perennial grasses was because of Themeda, while in unseeded plots it was because of a range of initially sparse species, particularly Austrostipa bigeniculata, Bothriochloa macra and Aristida ramosa. At Windermere, the increase was less substantial (18%, probably because native cover was already high; Table 2 and Fig. 4a) and effects differed between species. In particular, the warm-season C4 grasses Bothriochloa macra and Themeda australis increased while the C3 grasses were unaffected (Austrodanthonia spp.) or declined (Elymus scaber) under carbon supplements (Table 2).

Burning generally led to a decline in exotic annual grasses and an increase in broad-leaf annuals by the following spring (Table 2). Density and size data indicated that the decline in annual grasses was because of reduced density (P < 0·001), whilst broad-leaf species such as Echium plantagineum and Trifolium spp. increased in density and sometimes size (P < 0·001). The net outcome of these opposing effects of burning on exotic annuals differed between sites (Table 2 and Fig. 4). Total exotic abundance declined at Green Gully, although this decline was significantly smaller than on carbon-supplemented plots (Table 2). At Windermere, exotic grasses were severely diminished across all treatments in 2003 because of drought, thus the increase in broad-leaf annuals (particularly Trifolium subterraneum and other legumes) on burnt plots resulted in an overall increase in exotic annual abundance because of burning (Table 2 and Fig. 4a).

Burning often led to enhanced abundance of native perennial grasses by April 2004. At Green Gully this was significant only on plots with Themeda added (425% greater than untreated plots, interaction P < 0·001; Fig. 4b). At Windermere, dry conditions following the first burn led to reduced native grass abundance on burnt plots in spring 2003, but by autumn 2004 it had increased to 17% greater than control plots, despite the second burn (Table 2). Themeda australis and Digitaria hubbardii, both warm-season C4 species, contributed most to this increase, while C3 grasses such as Austrodanthonia spp. and Elymus scaber declined significantly because of burning (Table 2).

Given the low effectiveness of herbicide applications, the floristic composition of herbicide plots was generally similar to control plots.

perennial grass establishment

As noted earlier, no Poa establishment was recorded. Themeda established despite extremely dry conditions, and by April 2003 establishment was greatest on burnt plots, intermediate on carbon-supplemented plots and low to negligible on control and herbicide plots (Fig. 5). Dry conditions led to moderate mortality of Themeda seedlings, especially on burnt plots that had no protective litter layer. Thus by April 2004 Themeda density on burnt plots had declined to levels similar to carbon-supplemented plots at Windermere (Fig. 5a). At Green Gully, herbicide application appeared to reduce Themeda establishment or survival (Fig. 5b). Themeda abundance in April 2004 showed similar patterns to seedling numbers (Table 2).

Figure 5.

Plot treatment effects on Themeda establishment, for seeded plots only. Significance levels (*P < 0·05; **P < 0·01; ***P < 0·001) refer to comparisons within measurement dates; different letters indicate significant differences between values within measurement dates (P < 0·05).

As noted earlier, addition of Themeda seed led to increased native perennial grass abundance on burnt and carbon-supplemented plots at Green Gully. At Windermere, Themeda addition led to significantly greater (P < 0·04) perennial grass abundance across all treatments, although these effects were not apparent until April 2004 (Fig. 4).

effects ofthemedaon soil nutrients

At both sites, Themeda establishment influenced soil nitrate levels (Fig. 6). Remarkably, soil nitrate was significantly lower in Themeda plots at Windermere in March 2003 (P = 0·03) within 4 months of Themeda establishing in high densities. This effect was also detected during the second nitrate peak in April 2004, despite a poor growing season and locust damage. By this date, nitrate was significantly lower only on burnt plots with Themeda (interaction P= 0·04; Fig. 6a). This interaction was not unexpected, as control plots had low Themeda establishment and carbon-supplemented plots already had low soil nitrate. At Green Gully a similar effect of Themeda on soil nitrate was beginning to emerge by April 2004 (P = 0·057). Although the sowing–plot treatment interaction was not significant, the decline in nitrate was most strongly evident on burnt and carbon-supplemented plots (Fig. 6b), which had the highest abundance of Themeda (Table 2).

Figure 6.

Seed treatment effects on soil nitrate within each plot treatment, across the experimental period. Significance levels (*P < 0·05; not shown, not significant) refer to comparisons within measurement dates across all plot treatments. For significant seed effects, Themeda plots had significantly lower nitrate than other plots, except for April 2004 at Windermere where the significant interaction indicates significantly lower nitrate on Themeda plots within burn treatments only.

The magnitudes of changes to soil nitrate on Themeda plots were not as great as those induced by carbon supplements. Nevertheless, for burnt plots the reduction was still substantial, with nitrate levels on Themeda plots reduced by a mean of 46% at Windermere and 59% at Green Gully by the final measurement date.

There were no significant effects of seed treatments on any other soil nutrients, except for a seed–treatment interaction for ammonium at Windermere in March 2003 (Fig. 2a). The latter result was ambiguous, indicating significantly lower ammonium on Themeda plots for comparisons within burnt plots (consistent with effects on nitrate) but significantly higher ammonium on Themeda plots for comparisons within untreated plots.

Discussion

spring burning

Effects of fire on soil nutrients are variable and depend on fire temperature regimes, the amount and quality of residues, modifications to microclimate, and longer-term vegetation changes (Raison 1979). Available nutrients often increase temporarily after a single fire (as occurred after our first spring burns) because of processes such as direct release from mineral soil and organic matter, deposition in partly combusted residues and increased microbial activity (Raison 1979; Ojima et al. 1994).

Repeated burning over longer periods can deplete soil nutrients, particularly nitrogen, through repeated volatilization and changes to plant nutrient composition (Seastedt & Ramundo 1990; Ojima et al. 1994; Marrs 2002). This in turn can lead to lower nitrogen availability (Raison 1979; Seastedt & Ramundo 1990; Ojima et al. 1994). For example, soil-available and organic nitrogen declined significantly within 2 years of implementing annual burns in tallgrass prairies (Ojima et al. 1994). We detected a decline in soil ammonium (Windermere only) and no nitrate flush after our second fire, but it is unclear whether the different response in the second year was because of differing seasonal conditions or whether it reflected longer term nutrient losses. Longer term losses in total carbon and nitrogen were potentially beginning to occur at both experimental sites (Table 1) but these trends were not significant.

Spring burns are thus unlikely to be useful for reducing soil-available nitrogen in the short term. While repeated burns may lead to lower available nutrient levels, the longer time frames required are disadvantageous. In addition, depletion of total nutrient pools is not necessarily desirable, as total soil nitrogen is only weakly related to invasion by exotic annuals in grassy white box woodlands (Table 1; Prober, Thiele & Lunt 2002). Conversely, the minimal declines noted in the measured nutrients after two spring burns and historical adaptation to occasional fire in these ecosystems (Lunt & Morgan 2002) indicate that judicious, short-term, use of burning for other purposes (see below) should not be detrimental to soil nutrient stores.

Burning led to a substantial decline in annual grasses and an increase in broad-leaf annuals, consistent with Prober, Thiele & Koen (2004). Given the irregular effects of burning on soil nutrients, it is likely that effects on annuals were related to other ecological effects of burning. Many of the annual grasses involved have transient to short-lived seed banks compared with longer-lived seed banks in the broad-leaf annuals (Russi, Cocks & Roberts 1992; Morgan 1998; Grigulis et al. 2001), suggesting seed bank depletion led to reduced annual grass density. Furthermore, germination of hard-seeded species such as legumes may have been enhanced by fire (Bell, Plummer & Taylor 1993) and modified establishment conditions may have favoured broad-leaf annuals over annual grasses (Heady 1956). Unfortunately the herbicide treatment, designed to help distinguish among these possibilities, was unsuccessful.

Irrespective of effects on soil nutrients and exotic annuals, spring burns effectively enhanced native abundance on plots seeded with Themeda (Fig. 4). This increase was largely because of superior Themeda establishment after burning, which may have resulted from effects on seed dormancy (Cole & Lunt 2005) and establishment conditions. As Themeda established after annuals had died, enhanced establishment may have been because of the removal of the dense litter layer (but see Cole & Lunt 2005) rather than reduced competition from annuals.

Where Themeda was not added, increases in native abundance because of burning were smaller and not always significant. At Green Gully it is likely there were insufficient native propagules to replace annual grasses. Results at Windermere suggest that success of spring burning will be lower if relying on existing native species and/or propagules (e.g. Bothriochloa and Austrodanthonia spp.) for replacing exotics, as we found in an earlier study (Prober, Thiele & Koen 2004).

These results indicate that spring burns will be most effective for weed control where exotic annual grasses are the most problematic weeds (cf. perennials or broad-leaf annuals) and for enhancing native cover when combined with supplementary seeding of native grasses. Additionally, spring burns may be more effective on low nutrient sites, as broad-leaf annuals increased more dramatically after burning at Windermere and Green Gully than in our earlier study (a low nutrient site; Prober, Thiele & Koen 2004). For broader scale application, spring burns would be best applied in patches to minimize soil erosion and other undesirable impacts. Other techniques that remove biomass before annuals set seed, such as pulse grazing or slashing, might also be effective (Menke 1992).

carbon supplements

Carbon supplements reduced soil nitrate by up to 27 mg kg−1, to very low levels comparable with those recorded in little-disturbed white box woodlands with high-quality Themeda–Poa understoreys (0·13–1·8 mg kg−1; Prober, Lunt & Thiele 2002). Significant and rapid effects of carbon on soil nitrate are consistent with other studies, although few have recorded changes of such magnitude (Jonasson et al. 1996; Zink & Allen 1998; Reever Morghan & Seastedt 1999; Török et al. 2000; Blumenthal, Jordan & Ruselle 2003). The increase in soil nitrate on carbon-supplemented plots at Green Gully in April 2004 suggested that carbon supplements suppressed soil nitrate for approximately 3 months. This is consistent with expectations for labile carbon sources such as sucrose (Blumenthal, Jordan & Ruselle 2003) and confirms the need for regular supplements until soil nitrate is stabilized at desirable levels by other means (see below).

Carbon supplements affected soil properties other than soil nitrate, but to much lesser extents. Declines in soil ammonium were comparable with other studies (Reever Morghan & Seastedt 1999; Corbin & D’Antonio 2004) and led to levels similar to those recorded in little-disturbed woodlands (1·5–4 mg kg−1; Prober, Lunt & Thiele 2002). Occasional effects on soil phosphorus (Fig. 3) probably reflect an increase in phosphorus use by soil microbes (Jonasson et al. 1996). Phosphorus remained at higher levels than recorded in little-disturbed woodlands (1–3 mg kg−1; Prober, Lunt & Thiele 2002) and the influence of more substantial phosphorus reductions (e.g. through addition of ferric sulphate; Marrs 2002) on floristic composition would be of interest. As expected, total soil carbon increased with carbon supplements, but increases in total soil nitrogen, as sometimes observed in other studies (Zink & Allen 1998), were only marginally significant.

Consistently low soil nitrate levels resulting from carbon supplements (perhaps in combination with more minor reductions in ammonium and phosphorus) were highly effective in reducing exotic annuals. In contrast with spring burns that affected plant density, carbon supplements decreased plant growth rates. These results substantiate predictions from soil surveys that low soil nitrate in little-disturbed grassy woodlands limits the growth of exotic annuals (Prober, Thiele & Lunt 2002).

Reduced growth of exotic annuals enhanced the abundance of native species and dramatically increased germination and establishment of the target species Themeda. A number of studies have used carbon supplements to reduce undesirable species (Reever Morghan & Seastedt 1999; Alpert & Maron 2000) but few have successfully enhanced desirable species (Zink & Allen 1998; Paschke, McLendon & Redente 2000; Blumenthal, Jordan & Ruselle 2003; Perry, Galatowitsch & Rosen 2004). For carbon supplements to enhance native species, exotics must be more nitrophilic than natives (Blumenthal, Jordan & Ruselle 2003) and must significantly limit their establishment or growth (Lenz 2004). These conditions appear to be fulfilled in white box woodlands, although trends at Windermere indicate that some native species are disadvantaged by low nutrient levels (particularly Elymus scaber, a short-lived native perennial grass; Table 2). These trends are consistent with more general observations that high nitrate levels favour annual and early seral species, while perennial grasses can efficiently utilize nitrate at low concentrations (Muller & Garnier 1990).

While carbon supplements are a potentially effective tool for restoring grasslands, their on-ground application requires considerable refinement. Efficiencies might be achieved by determining minimum effective rates and optimal timing of carbon applications for sites of differing fertility, and investigating the use of cheaper and less labile carbon sources such as sawdust.

restoring ecological function in grassy woodland understoreys

In the long term, management interventions such as carbon supplements and spring burns will only assist restoration if they help establish target ecosystems that resist further weed invasion (Bakker & Wilson 2004). Woodland remnants will always be subject to exotic seed rains from surrounding landscapes. Seed influx may be reduced through landscape design (e.g. buffer zones) and propagule accumulation may be limited by occasional spring burns or related techniques (Menke 1992; Prober, Thiele & Koen 2004). However, ensuring that conditions are unfavourable to such species potentially provides the strongest mechanism for avoiding repeat invasion.

Targeting soil nitrogen shows particular promise for increasing resistance to invasion by nitrophilic exotics. Nitrogen cycling in ecosystems is almost exclusively regulated by biological processes (Török et al. 2000), and an effective point of intervention is addition of carbon to enhance nitrogen use by soil decomposing organisms and thus reduce nitrogen availability to plants. This change is only temporary, however, and is best viewed as creating a window of opportunity for encouraging more persistent biological changes that maintain lower available nitrogen in the long term.

Such longer term changes include modified interactions between the vegetation and nitrogen cycling that result from altered vegetation composition. Earlier studies have indicated that annual-dominated ecosystems persist through positive feedback mechanisms that ensure high nitrogen supply during critical growth phases (Vitousek 1982; Jackson et al. 1988; Prober, Thiele & Lunt 2002). Perennial systems, in contrast, can maintain lower available nitrogen throughout the year through a higher capacity to extract soil nitrate, more extensive root systems with year-round activity, and production of low-nitrogen litter (Tilman & Wedin 1990; McLendon & Redente 1992; Redente, Friedlander & McLendon 1992). Replacing annuals (or other nitrophilic exotics) with appropriate native perennial species may thus provide a key to restoring stable, low-nitrate conditions in degraded grassy woodlands. Experimental evidence showing that Themeda establishment can reduce soil nitrate levels (Fig. 6) is consistent with this prediction.

The rapid rate at which Themeda plants reduced soil nitrate suggests these initial effects occurred through mechanisms such as a greater capacity to extract soil nitrate. From the April 2004 results it is clear that Themeda activity was insufficient to prevent resurgence of soil nitrate on carbon-supplemented plots at Green Gully, or to reduce soil nitrate to inferred pre-European levels (Prober, Lunt & Thiele 2002) on burnt plots. However, Themeda plants were small and there remains considerable scope for greater suppression of soil nitrate as plants mature. Further studies are needed to confirm this prediction, and whether this could contribute to the additional goals of reducing exotics and enhancing establishment of other native species.

It is notable that Themeda reduced soil nitrate at Windermere even though other native perennial grasses already dominated these plots. This could reflect the small overall increase in native perennials on these plots by autumn 2004 (Fig. 4), or Themeda may have a greater effect than other native grasses on soil nitrate. The latter is supported by evidence for consistently lower nitrate levels beneath Themeda swards than beneath other native grass swards (Prober, Thiele & Lunt 2002) and evidence that perennial grass species can differ greatly in their effects on soil nitrate (Tilman & Wedin 1990). Studies comparing effects of relevant native perennial grasses on soil nutrients are thus needed to guide selection of species for restoring ecosystem function in temperate Australian woodland. Of particular interest is the influence of Poa, once codominant with Themeda in woodland understoreys; this cool-season C3 perennial may influence soil nitrogen during winter and early spring, when Themeda is dormant and exotic annuals are active. The contribution of subsidiary forbs and grasses to this process also requires assessment.

Given the potential significance of Themeda (or other natives that suppress available nitrogen) for restoring ecosystem function, management interventions that promote rapid establishment of a dense Themeda sward may be critical for maximizing outcomes in understorey restoration projects. While they acted through different mechanisms, spring burning and carbon supplements both promoted Themeda establishment and growth. Indeed, these interventions may be complementary, and an efficient management strategy may involve spring burns to encourage dense Themeda establishment, followed by carbon supplements to suppress exotics and promote rapid Themeda growth.

Previous authors have recommended re-establishing Themeda as a principal restoration goal in grassy woodlands of south-eastern Australia (Cole & Lunt 2005), largely because it was a dominant species in pre-disturbance understoreys. However, few studies have demonstrated why Themeda should play a special role. Our results indicate that Themeda may be a keystone species, able to drive and maintain the soil understorey system in a low-nitrate condition that, if appropriately managed, remains resistant to weed invasion. Such an outcome fulfils a central goal of restoration activities.

Acknowledgements

We thank the Young Rural Lands Protection Board and the Johnson family for allowing us to conduct trials on their sites, and Isobel Crawford, Ray Dowling, Hugh Jackson and Kerry Seaton for field assistance. This study was supported by the New South Wales government through its Environmental Trust.

Ancillary