1
1
Does long-term fire exclusion in an Australian tropical savanna
2
result in a biome shift? A test using the re-introduction of fire.
3
4
5
Kenneth ScottA, Samantha A. SetterfieldA*, Michael M. DouglasA,B, Catherine L. ParrC,
Jon SchatzD, and Alan N. AndersenD.
6
7
A
8
Darwin, NT, 0909, Australia.
9
B
Research Institute for Environment and Livelihoods, Charles Darwin University,
Tropical Rivers and Coastal Knowledge Research Hub, Charles Darwin University,
10
Darwin, NT, 0909, Australia.
11
C
12
Environment, Oxford OX1 3QY, England
13
D
14
0822, Australia.
University of Oxford, Environmental Change Institute, School of Geography &
CSIRO Ecosystem Sciences, Tropical Ecosystems Research Centre, Winnellie, NT,
15
16
*
Corresponding author:
17
Dr Samantha Setterfield
18
Research Institute for the Environmental and Livelihoods
19
T. +61 8 8946 6756 | F. +61 8 8946 6847 | M. 0427 456 031
20
[email protected]
21
22
23
Running Title: Re-introducing fire to an Australian savanna
2
1
Abstract
2
The structure of tropical savanna ecosystems is influenced by fire frequency and
3
intensity. There is particular interest in the extent to which long-term fire exclusion can
4
result in a shift from savanna to forest vegetation that is not easily reversed by the re-
5
introduction of fire. This study examined changes in the structure and composition of a
6
long-unburnt site within the northern Australian savannas following a long period of
7
active fire exclusion (>20 years), and the effect of the re-introducing fire through
8
experimental fire regimes, including fires in the early and late dry season at a range of
9
frequencies. After the long period of fire exclusion, the vegetation community was
10
characterised by a well developed mid-story and canopy layer, low grass cover,
11
substantially higher densities of woody sprouts and saplings than frequently burnt
12
savanna. The community composition included a high proportion of rainforest-affiliated
13
species. Three years of experimental fires had no detectable effect on the overall
14
composition of grass-layer and woody plants but had an effect on woody vegetation
15
structure. Continued fire exclusion further increased the density of woody stems,
16
particularly in the mid-storey (2.0-4.99 m), whereas moderate-intensity fires (> 800 kW
17
m-1) significantly reduced the density of mid-storey stems. The re-introduction of higher
18
moderate intensity fire events resulted in the vegetation in some compartments reverting
19
to the open savanna structure typical of frequently burnt sites. Such rapid reversibility
20
suggests that in general, the woody thickening resulting from long-term fire exclusion
21
did not represent a biome shift to a non-savanna state. However, there was a small
22
proportion of the site that could not sustain the fires applied to them because grass cover
23
was very low and patchy and therefore appeared to have crossed an ecological threshold
24
towards closed forest.
25
Keywords. savanna, fire, biome shift, woody thickening, tropical woodland
3
1
2
INTRODUCTION
State-and-transition models are particularly useful for describing the range of
3
complex ecosystem processes and multiple vegetation states observed in tree-grass
4
ecosystems (Westoby et al. 1989; Angassa & Oba 2008). According to state-and-
5
transition theory, vegetation can shift from one stable state to another, and this shift can
6
be irreversible without dramatic intervention (Briske et al. 2003; Stringham et al. 2003;
7
Bestelmeyer et al. 2003; Briske et al. 2006). Substantial changes in composition and
8
structure may occur within a vegetation state, but negative feedbacks reinforce
9
ecosystem stability, and changes are easily reversed by removing the stressor/s that
10
11
drove the initial change (Bestelmeyer et al. 2003; Briske et al. 2006).
In extreme cases a transition from one state to another can involve such a
12
transformation in ecosystem structure and function that it represents a shift from one
13
biome to anther. This is the case throughout the seasonal tropics, where savanna and
14
forest represent contrasting biomes in regions of high rainfall (Boulière 1983; Bond
15
2008; Bond & Parr 2010). Savanna vegetation is dominated by fire-resistant, shade-
16
intolerant, sclerophyll tree species that form an open canopy over a continuous grass
17
layer, whereas forest is dominated by fire-sensitive, shade-tolerant mesophyll tree
18
species that form a closed canopy with little or no grass cover (Ratnam et al. 2011;
19
Thompson et al. 2001; Wilson et al. 1990). Long-term fire exclusion sees a transition
20
from savanna to forest, as juvenile woody plants to recruit into the midstorey layer, and
21
the increased shade reduces grass biomass (Brookman-Amissah et al. 1980; San Jose &
22
Farinas 1991; Silva et al. 2001; Russell-Smith et al. 2003; Woinarski et al. 2004;
23
Williams RJ et al. 2003). The reduction in grass biomass results in reduced fire
4
1
frequency, which reinforces the transition from savanna to forest (Ratnam et al. 2011;
2
Scholes & Archer 1997; Russell-Smith et al. 2003; Sharp & Whittaker 2003).
3
Alternatively, frequent burning has been shown to reverse woody thickening,
4
particularly in combination with drought and browsing by large herbivores
5
(Freudenberger et al. 1997; Roques et al. 2001; Augustine & McNaughton 2004; Holdo
6
2007, Midgley et al. 2010). Most savanna woody species are capable of re-sprouting
7
after fire, although high mortality occurs in response to higher-intensity fires,
8
particularly of the smallest and largest individuals (Midgley et al. 2010; Setterfield
9
1997; Williams RJ et al. 2003; Werner 2005). Conversely, dry season burning
10
generally has little effect on grass-layer plants, with compositional change attributed to
11
inter-annual rainfall variability (O’Connor 1985; O’Connor & Everson 1998; Williams
12
RJ et al. 2003; Williams PR et al. 2003b). Site and stand characteristics, and the
13
interaction between these, may alter these responses because they determine fire
14
behaviour and the response of the vegetation community to fire behaviour (Williams et
15
al. 2003; Bowman & Fensham 1991). Fensham et al. (2003) demonstrated a major (78-
16
89%) reduction in tree basal area following burning of a fire-protected savanna, where
17
rainforest-associated species and obligate seeders were particularly susceptible to
18
repeated burning.
19
Open savanna communities characterised by a continuous understorey of C4
20
grasses and a scattered overstory of Eucalyptus species dominate northern Australia
21
(Setterfield and Hutley 2008). These communities are frequently burnt every 1-3 yrs
22
(Andersen et al. 2005). Within the savanna matrix are small (typically <5ha) pockets of
23
closed monsoon rainforest (Russell-Smith 1993; Russell-Smith and Setterfield 2006). A
24
range of studies in the savannas of northern Australia have documented transition
5
1
towards closed forest in the long-term absence of fire, characterised by increases in
2
stem density, development of the midstory or subcanopy, decrease in grass cover and
3
increase in rainforest-associated species (Russell-Smith et al. 2003; Williams PR et al.
4
2003a; Woinarski et al. 2004). However, it is unclear at what point along this transition
5
the shift from savanna to forest can be considered to have occurred. One approach to
6
addressing this is to focus on the ease with which vegetation change is reversible. If the
7
re-introduction of fire results in a rapid reversion to open savanna then it can be argued
8
that a biome shift has not yet occurred.
9
Here we use this approach to assess if >20 years of active fire exclusion at a
10
savanna site in northern Australia has resulted in a biome shift. First, we describe the
11
state of the vegetation in the long-term absence of fire, with the expectation that it will
12
have characteristics of forest structure, composition and function, such as a substantial
13
proportion of forest species, a relatively closed canopy, and patchy grass cover (and
14
therefore a low capacity to carry fire). We then assess the vegetation response to the re-
15
introduction of fire focussing on the extent to which these forest characteristics are
16
reversed.
17
METHODS
18
Study area
19
The study was undertaken within the Territory Wildlife Park (TWP) at Berry
20
Springs, 30 km southeast of Darwin, Northern Territory, Australia (12º 41′ 42.25″ S,
21
130º 58′ 50.36″ E). The region experiences a monsoonal tropical climate, with a hot
22
summer wet season between October and March, and a mild winter dry season between
23
April and September. Mean annual rainfall at the Middle Point Automatic Weather
6
1
Station (36 km to the north-east) is 1399 mm, of which 90% falls in the wet season
2
(Bureau of Meteorology, unpublished data, 2007). Rainfall recorded at the Park during
3
the study (2004-2007) was 17% higher than the long-term average at Middle Point,
4
principally as a result of a cyclone-related rainfall event in April 2006 (Bureau of
5
Meteorology and L Smith, unpublished data, 2007).
6
Fire has been actively excluded from the Park since at least 1981. Only two
7
patchy fires had occurred in the study site between 1981 and the application of
8
experimental fire regimes in 2004. In 1980, vegetation structure at the site was typical
9
of the broader region; a sparse midstorey (approximately 2-5 m high) over a grass-layer
10
of tall perennial (Heteropogon triticeus) and annual (Sorghum intrans) grasses
11
(Sivertsen et al. 1980). Eucalyptus tetrodonta F.Muell. and Eucalyptus miniata
12
A.Cunn. ex Shauer were dominant overstorey trees (Sivertsen et al. 1980; Fig. 1). At
13
the beginning of this study (2004), a moderately- to well-developed midstorey was
14
present throughout the site (Fig. 1), particularly in association with the Quaternary
15
colluvial sediments of the permanent lagoon in the south (Scott et al. 2009). Common
16
midstorey shrubs included Calytrix exstipulata DC., Planchonia careya (F.Muell.)
17
Kunth, Petalostigma pubescens Domin and Exocarpus latifolius R.Br. Common
18
herbaceous plants in the grass-layer were low-biomass species, such as the annual grass
19
Pseudopogonatherum contortum (Brongn.) A.Camus, sedge Fimbristylis sp. Charles
20
Darwin (J.L.Egan 5300) and perennial grass Eriachne triseta Nees ex Steud. (Scott et
21
al. 2009).
22
Experimental Design
23
The fire experiment at the TWP consisted of a line of 18 1-ha compartments
24
located in the north-western corner of TWP, arranged in three contiguous blocks (North,
7
1
Centre and South) to accommodate the environmental variability at the site. A detailed
2
description of the variation is provided in Scott et al. (2009). In summary, soil moisture
3
content of the surface soil was significantly higher in the North and Central blocks
4
where the soils are gravelly clay loams on a small ridge, and exhibit an horizon of
5
massive laterite of Tertiary origin, 10–50 cm below the surface (Scott 2008). In the
6
South block, soils were sandy loams associated with Quaternary colluvial sediments of
7
a permanent lagoon, significantly less gravelly (25% gravel >2mm in the upper 5 cm
8
soil surface), positioned at the foot of the ridge, and did not contain massive laterite in
9
the upper 1m of the soil profile (Scott 2008). Vegetation structure varied along the
10
edaphic gradient, with significantly higher density of woody stems in the midstorey in
11
the South block (~2000 stems ha-1) compared with the North and Central Blocks (1000-
12
1100 stems ha-1).
13
Fire regimes were allocated as treatments to three compartments each, one per
14
block, according to a randomised complete block design. This allowed variation in
15
edaphic characteristics to be assessed in the experiment. Fire regimes were: unburnt
16
(UN), annual early fires (Annual Early), a single early season fire in 2004 (Early 04), a
17
single early season fire in 2005 (Early 05), and biennial late (Biennial Late). All fire
18
regimes were first applied in 2004, with the exception of the Early 05 regime which was
19
first burnt in 2005 (Table 1). Experimental fires were lit with drip torches along the
20
leeward sides of each compartment to be burnt (back fire). After the fire had
21
established within the compartment, fire lines were lit on the windward side unless the
22
back fire was intense and moving rapidly to the windward side. This paper reports on
23
changes in vegetation in the first three years following the reintroduction of fire. The
24
early season fires were typically lower intensity, and therefore lower flame height, than
8
1
late season fires which reflects the fire behaviour in this region (Table 1; Andersen et al.
2
2005; Setterfield et al. 2010). The exception was the Early 05 treatment which had a
3
moderate intensity fire, and contrasted with the early fires in other treatments. There
4
was variability within fire regimes due to the edaphic and vegetation differences
5
between compartments and blocks. For example, very low fire intensities occurred in
6
some compartments, particularly in the South block, where dense woody vegetation and
7
high canopy cover resulted in sparse grass cover (Fig 1b, Scott et al. 2009).
8
Plant composition and abundance
9
To determine if the reintroduction of fire caused a substantial change in structure
10
and composition of woody plants, surveys were completed in May 2004, before fire
11
regimes were first implemented, and again in March 2007. Sampling was undertaken
12
within two parallel belt transects (80 × 6 m, approximately 30 m apart) within each
13
compartment, where the abundance of woody plant species (including standing dead
14
stems) was recorded. For stems < 2 m, abundance was categorised according to two
15
height strata: sprouts (0-0.49 m) and small saplings (0.5-1.99 m) (collectively known as
16
the grass-layer stratum). For stems > 2 m tall (and > 3 cm dbh), the height and dbh of
17
each individual was recorded, which enabled categorisation into the following height
18
strata: lower midstorey (2.0-4.99 m), upper midstorey (5.0-9.99 m) and overstorey (10.0
19
m+). Stems > 2m, but < 3 cm dbh, were regarded as large saplings. The precise height
20
of large saplings was not measured, although they were typically < 5 m (i.e. in the lower
21
midstorey), and contributed only a very small proportion (2.4%) of the total woody stem
22
density (and therefore basal area). Basal area (m2 ha-1) was calculated by multiplying
23
stem area (obtained from dbh measurements) by stem density, considering each woody
9
1
stem > 2 m height and > 3 cm dbh. Eight uncommon vine species were present but
2
were not included in the site assessment.
3
Most savanna woody plants can regenerate vegetatively following fire. To
4
document the extent of structural change associated with stem mortality and subsequent
5
basal sprouting (where no mortality takes place), stems in the 2007 survey were noted if
6
this response had occurred (individuals are termed ‘basal sprouters’).
7
To assess the effect of fire regimes on the grass-layer species, the abundances of
8
grass, sedge and forb species were sampled annually in the late wet season from 2004 to
9
2007. The density of each species was counted within 18 2-m2 permanent quadrats per
10
compartment; nine quadrats were spaced, every 10 m, along the centre line of each
11
swathe established to sample the woody vegetation. To aid the interpretation of floristic
12
responses to fire, tree litter and canopy cover were also assessed annually at these
13
quadrats. The percent cover of grass and leaf litter was estimated visually (to the nearest
14
5%), and percent canopy cover recorded using the mean of four cardinal points on a
15
forestry densitometer immediately outside each quadrat (Lemmon 1957).
16
Sampling occurred in the late wet season (March-May) within a narrow window
17
of opportunity when the ability to detect and identify grass-layer plants was maximised,
18
given the presence of mature, flowering individuals. A slightly later sampling period in
19
2004 made for unreliable identification of annual species (grasses, forbs and sedges),
20
because they had already set seed and begun to cure. With the exception of perennial
21
grasses and selected annual grass species, most analyses therefore consider the surveys
22
of 2005 to 2007 inclusive.
10
1
For some other plants, identification to species level was not possible given the
2
absence of flowering material during the sampling period. Such species were included
3
as aggregates of two or more species with a similar vegetative appearance: Eriachne
4
agrostidea F.Muell included E. ciliata R.Br.; Panicum mindanaense Merr. included
5
Yakirra nulla Lazarides & R.D. Webster and Urochloa polyphylla (R.Br.) R.D.
6
Webster; Eriachne triseta Nees ex Steud. included Aristida holathera Domain and E.
7
burkittii Jansen; and Eriachne avenacea R.Br. included E. major (Ewart & O.B.Davies
8
Lazarides).
9
10
Statistical analysis
Analysis of variance was used to assess changes in the abundance of woody
11
vegetation within each height stratum, species richness, and basal area. Data from
12
transects were averaged and the variables were analysed with multivariate ANOVA
13
with the following factors: Year (fixed), Regime (fixed), Block (random), with Block
14
nested in Regime. Variables were square-root or log-transformed if necessary, in this
15
and all other analyses, to satisfy the normality and variance homogeneity assumptions
16
of ANOVA.
17
Analyses of woody stem abundance were also conducted separately (in each
18
stratum) for the rainforest affiliated species (defined by Liddle et al. 1994 as species
19
that are “either exclusive to, or typically components of, perennially or seasonally
20
closed canopy vine-forest or vine-thicket vegetation”), given expected differences in
21
their responses to fire compared with savanna species (Fensham et al. 2003). Analyses
22
of woody stem abundance was also conducted for species based on the height of those
23
species when they are mature individuals (grass-layer species (<2m), 6 spp.; shrub
24
species (2 -10m) , 58 spp.; tree species (>10m), 12 spp.).
11
1
Two-way ANOVA was used to determine whether fire regime influenced the
2
incidence of basal sprouting. The number of woody stems exhibiting the basal
3
sprouting response (as a proportion of the total live stem density) was compared
4
between fire regime treatments in a design consisting of Regime (fixed), and Block
5
(random), with block nested in fire regime.
6
To make generalisations on grass-layer plant responses, species were
7
categorised according to life form: annual or perennial grasses (Poaceae), non-
8
leguminous forbs, sedges (Cyperaceae) and legumes (Fabaceae). Grass cover, species
9
richness (2 m2 scale), total density, and the density of each life form group, were
10
analysed with a repeated measures ANOVA to assess change over time within fire
11
regime treatments. Factors comprised Fire Regime (fixed factor), Block (random
12
factor), Transect (random factor) and Year (repeated measure, fixed factor), with
13
transect nested in fire regime and block.
14
Patterns of woody and grass-layer species composition at the site were analysed
15
separately with ordination and ANOSIM (analysis of similarity) using Primer 6 (Clarke
16
& Gorley 2001). The number of individuals within each sampling unit was converted to
17
a density value, and then averaged to the compartment level. Data were log (x+1) or
18
square root transformed, and resemblance matrices constructed using the Bray-Curtis
19
measure of similarity. One-way ANOSIM was used to determine whether species
20
composition differed between fire regime treatments and blocks at the end (2007) of the
21
sampling period.
22
Canopy, grass and litter cover prior to fire were compared between blocks using
23
one-way ANOVA. Tukey’s HSD posthoc tests were used to identify which blocks were
24
significantly different from each other. Litter and Canopy cover was compared between
12
1
fire regime treatments in each survey year (2004-2007 inclusive) using a non-parametric
2
Kruskal-Wallis Ranks test (as a result of positively skewed distributions with zero
3
counts).
4
5
RESULTS
6
The two woody vegetation surveys recorded 84 plant species at a mean density
7
of 9945 stems ha-1. Approximately half of the species (34 spp.), and a third of the stems
8
overall (36.1%), were rainforest-associated species. Rainforest species had a greater
9
prominence in the grass-layer strata, where they comprised 38.5% of the stems,
10
compared to 19.1% in the midstorey and 15.9% in the overstorey. The overwhelming
11
majority of woody stems were sprouts (73.6%), followed by small saplings (12.6%),
12
stems in the lower (7.1%) and upper (3.5%) midstorey, large saplings (2.4%) and
13
overstorey trees (0.7%).
14
Site characteristics varied along the edaphic gradient prior to the reintroduction
15
of fire. Canopy cover in 2004 was significantly lower in the North (58.6 ± 3.8%) and
16
Central (57 ± 2.1 %) blocks compared to the South block (80.6 ± 5.3; F2,15 =9.5,
17
P<0.01). Similarly litter cover in 2004 was significantly lower in the North (22.8 ±
18
2.5%) and Central (23.6 ± 4.5%) blocks compared to the South block (43.4 ± 3.9; F2,15
19
=9.62, P<0.01).
20
Edaphic conditions were clearly an important influence on vegetation with a
21
significant effect of block overall evident for abundance of sprouts, small saplings, and
22
stems in the upper midstorey and overstorey (Table 2). In all cases, abundance of these
23
groups was higher in the South block (sandy loams soils) than the North or Central
13
1
block (lateritic soils; Fig. 2). This trend was particularly evident for rainforest species
2
in these strata (Table 2).
3
Leaf litter cover increased at a similar rate between 2003 and 2007 in the Annual
4
Early, Early 04 and Unburnt regimes, demonstrating that lower fire intensities had little
5
effect on litter accumulation (Fig. 4). By contrast, the higher intensity fire in the Early
6
05 regime in 2006, and late dry season fires in the Biennial Late regime, markedly
7
reduced litter cover (Fig. 4). As a consequence, there was a significant difference in
8
litter cover between regimes in 2005 (H(4, n=30) = 15.25, P<0.01) and 2007 (H(4, n=30) = 17.34,
9
P<0.01), after some of the higher intensity fires.
10
Canopy cover remained relatively constant during the study period in the
11
Unburnt regime and in regimes with low-intensity or infrequent fires (i.e. Annual Early
12
and Early 04 regimes, Fig. 4). In contrast, canopy cover decreased (by approximately
13
15%) in the first year after the moderate-intensity Early 05 fire, and decreased (by
14
approximately 10%) in response to the second fire of the Biennial Late regime, the most
15
intense fire of the experiment, resulting in a significant difference in canopy cover
16
between regimes in 2007 (H(4, n=30) = 10.92, P<0.05).
17
Effect of fire on woody plant species
18
The reintroduction of fire altered stand structure through reduced abundance of
19
the large saplings and the lower midstorey stems, although the magnitude of this change
20
was determined by fire intensity resulting in Year by Regime interactions (Table 2).
21
The abundance of large sapling stems increased or remained relatively stable within
22
most plots, but decreased with the repeated moderate-intensity fires of the Biennial Late
23
regime (Fig. 5). The abundance of lower midstorey stems decreased within the fires of
14
1
the Early 05 and Biennial Late regimes but was maintained or increased within the UN,
2
and lower intensity Annual Early and Early 04 regimes (Fig. 5). A similar pattern
3
occurred for the abundance of the tree species, with the latter also demonstrating a
4
significant year by regime interactions (Table 2). The mean abundance of lower
5
midstorey stems overall (and tree species in that group) showed a significant effect of
6
year overall (Table 2). Stems in the upper midstorey strata and overstorey strata were
7
more resilient to burning than those in the lower midstorey, showing no effect of Year,
8
nor a Year x Regime interaction (Table 2; Fig. 5).
9
The fire regimes in this experiment had little effect on the sprout layer with
10
increases in mean sprout abundance evident in all fire regime treatments over time (Fig.
11
6); resulting in a significant Year effect. The significant increase was evident for the
12
rainforest, midstorey and overstorey groups in the sprout layer (Table 2). The
13
abundance of overstorey species in the sprout layer decreased in the higher-intensity
14
Early 05 regime in contrast to the increases observed within all other fire regimes,
15
resulting in a significant year by regime interaction (Table 2). The abundance of small
16
saplings also showed a significant increase over time, primarily due to the increase in
17
overstorey species (Table 2).
18
Despite the reduction in large saplings and midstorey stems, basal area was not
19
significantly affected by year, fire regime or block. However, there were significant
20
inter-annual changes in species richness and stem abundance. Species richness
21
increased significantly over time within all but the Biennial Late regime (Fig. 7), from a
22
mean of 24.7 species/transect in 2004, to 27.1 species/transect in 2007. There was no
23
significant effect of Fire regime or Block, or interaction between factors (Table 2).
15
1
Rainforest species showed a significant effect of block in all strata (Table 2),
2
reflecting the general pattern observed for the community (Fig. 3). Rainforest species in
3
the sprout and lower midstorey strata showed a significant effect of Year, whereby
4
sprouts increased considerably over time (particularly in the Annual Early and UN
5
regimes) and lower midstorey stems decreased, particularly in the higher intensity
6
regimes (Early 05 and Biennial Late; Fig. 8).
7
The abundance of standing dead stems tripled during the experiment, resulting
8
in a significant effect of Year and Block but no difference was detected between fire
9
regimes (Table 2). The mean number of dead stems was 38.5 stems ha-1 in 2004
10
(ranging from 82.3 ha-1 in the South block to 10.4 stems ha-1 in the North block), to a
11
mean of 160.8 dead stems ha-1 in 2007 (ranging from 259.4 stems ha-1 in the South
12
block to 59.4 stems ha-1 in the North block). Despite this increase, standing dead stems
13
(represented by 59 species) still only comprised a small proportion (1%) of the total
14
number of stems in 2007. The species with the highest proportion of standing dead
15
stems (with n > 10 individuals in total) were Calytrix exstipulata (Myrtaceae; 8.0%),
16
Acacia holosericea (Mimosaceae; 7.6%) and Acacia oncinocarpa (Mimosaceae; 7.3%)
17
in the grass-layer, and Grevillea pteridifolia (Proteaceae; 44.4%), Alphitonia excelsa
18
(Rhamnaceae; 42.2%) and Acacia latescens (Mimosaceae; 23.1%) in the midstorey /
19
overstorey.
20
In 2007, basal sprouting was evident on 59 species. Basal sprouting occurred on
21
7.4% of the woody stems, and the vast majority of basal sprouters (live tissue
22
component, rather than the often taller dead portion) were present in the grass-layer
23
(87.0%), some were present in the midstorey layer (12.8%), and only two individuals
24
were present in the overstorey layer (0.1%). There was a significant difference between
16
1
fire regime treatments in the proportion of individuals that exhibited basal sprouting (F
2
4, 10
3
the UN and the Early 04 regime (approximately 4 % of total stems in the regimes),
4
whereas basal sprouters comprised about 10% of stems in the Annual Early and Early
5
05 regimes, and 20% in the higher-intensity Biennial Late regime. The species for
6
which basal sprouters were most common (with n > 10 individuals in total) included
7
Lophostemon lactifluus (Myrtaceae; 43.8%), Syzygium eucalyptoides subsp. bleeseri
8
(Myrtaceae; 32.6%) and Corymbia bleeseri (Myrtaceae; 31.4%) for stems in the grass-
9
layer, and Syzygium eucalyptoides subsp. bleeseri (Myrtaceae; 40.0%), Acacia
= 6.98; P < 0.01; Fig. 9). The lowest proportion of basal sprouting was observed in
10
lamprocarpa (Mimosaceae; 28.1%) and Acacia mimula (Mimosaceae; 19.6%) in the
11
midstorey / overstorey.
12
Effects on grass and legumes species
13
In total, 116 herbaceous grass-layer plant species were identified during the
14
annual surveys, at a mean density of 70 individuals m-2 (range: 0-1802 m-2). The grass-
15
layer was dominated by annual grasses (36.9% of total density, 17 spp.) and annual
16
forbs (28.9%, 56 spp.). Sedges (14.9%, 11 spp.), perennial grasses (13.4%, 18 spp.) and
17
legumes (5.9%, 14 spp.) together comprised the remaining one-third of the grass-layer
18
plants. The most common species (according to density) were the annual grass
19
Pseudopogonatherum contortum (Poaceae; 11.7 ± 1.1 SE individuals per m-2), sedge
20
Fimbristylis sp. Charles Darwin (Cyperaceae; 7.0 ± 0.7 per m-2) and perennial grass
21
Eriachne triseta (Poaceae; 6.1 ± 0.3 per m-2).
22
Grass cover was low at the commencement of the fire experiment (12.9 ± 2.2
23
%), and with a small overall increase to 18.0 ± 2.3 after the three years for experimental
24
fires). There was no significant increase in grass cover was in the higher intensity
17
1
Biennial Late regime, where cover increased from 11.7 ± 7.1 % in 2004 to 27.9 ± 3.1
2
%). However, the total density of individuals in the grass layer, and the density of
3
annual grasses and legumes specifically, showed a significant effect of year overall,
4
with density increasing over time (Table 3). The mean density of all life forms, except
5
sedges, progressively increased in each year of the experiment (Table 4). For both
6
annual grasses and total density, a significant year by regime interaction was not evident
7
(Table 3), indicating that these increases occurred in spite of the different fire regime
8
treatments applied. Total grass-layer density and annual grass density increased by ~
9
45% and 100% respectively over the sampling period. The increase in density during
10
the experiment coincided with an increase in the total rainfall and number of rain days
11
during the wet season (November to February inclusive) prior to sampling (from 2005
12
onwards; Table 4). The increase in grass-layer plant density also coincided with a
13
decrease in the total rainfall and number of rain days in the late dry season (August to
14
October inclusive) prior to sampling. Species richness and the density of all life form
15
groups showed a significant effect of block overall (Table 3).
16
Perennial grasses and legumes showed a significant year by regime interaction,
17
with changes in density over time determined by the fire regime treatments imposed
18
(Table 3). Perennial grass density increased gradually in the UN regime, to double
19
within three years. A similar pattern occurred within the Early 04 regime which was
20
burnt early in the experiment but protected from fire thereafter (Fig. 10 a). Regular
21
burning and regimes with higher-intensity fires tended to limit population increases;
22
perennial grass density in such regimes (e.g. Biennial Late) remained relatively stable
23
during the experiment.
18
1
The significant regime × year interaction of legume density was the result of a
2
substantial increase in density in the first year after fire, particularly after the higher-
3
intensity fires of the Early 05 and Biennial Late regimes (Fig. 10 b). The post-fire
4
response was smaller after lower-intensity fires of the Annual Early regime, and no
5
increases were evident in the UN regime. Indeed, density tended to decrease after the
6
initial post-fire increases (e.g. Fig. 10 b, Early-05, 2006-2007).
7
Density of the annual grass Sorghum intrans was low (≤ 0.06 individuals per
8
m2) before burning commenced and remained so three years after the application of fire
9
regime treatments. There was, however, a notable increase in abundance towards the
10
end of the experiment, apparently unrelated to fire regime (Fig. 10 c).
11
Floristic composition
12
After three years, there was no significant difference in woody (R = -0.079; P =
13
0.721) or grass-layer (R = -0.119, P = 0.756) species composition between fire regime
14
treatments. At this time, however, there was a strong, significant effect of Block
15
evident for both components, with the open forest community in the North and Central
16
Block separating from the South Block (woody plants, R = 0.369, P = 0.002; grass-layer
17
plants, R = 0.606, P = 0.002). This dissimilarity of the woody component was primarily
18
due to the differences in midstorey taxa (Table 5), with five of the six species that
19
contributed more than 3% towards the dissimilarity being midstorey species, and the
20
other being a canopy tree. Four of these species (including three rainforest species) were
21
most common in the ecotone community of the South Block. Most grass-layer plant
22
species that contributed to the dissimilarity in composition between the woodland
23
(North and Central Blocks) and ecotone (South Block) were more abundant in the
19
1
woodland, with the notable exception of E. triseta, which was much more abundant in
2
the ecotone (Table 5).
3
DISCUSSION
4
Vegetation structure after long-term fire exclusion
5
The high density of woody plants at the site at the commencement of the study,
6
particularly in the lower midstorey, suggests that a significant change in vegetation
7
structure occurred during the period of fire exclusion. The density of woody sprouts
8
and small saplings at the site (500 – 1000 stems/ha) was considerably higher than
9
typically occurs in Australian savannas (Bowman & Panton 1993; Bowman & Panton
10
1995; Vigilante & Bowman 2004). In contrast, woody stem density was similar to that
11
at another Australian savanna site (Munmarlary; 1400 midstorey stems/ha) where fire
12
had been excluded for >20 years, and where midstorey stems were previously absent
13
(Russell-Smith et al. 2003). At our site, edaphic characteristics strongly affected the
14
development of the midstorey, with significantly higher midstorey stem density in the
15
South block, where the sandy loam soils with low gravel content are likely to result in
16
higher plant available moisture compared with the North and Central blocks where the
17
soils are clay loams with high gravel content and a shallow lateritic layer (Scott et al.
18
2009). The South block is located close to a permanent freshwater lagoon, and therefore
19
the watertable is likely to be higher throughout the dry season than on the ridge in the
20
north (Scott et al. 2009). This process of woody thickening is also supported by the
21
trends documented in the unburnt regime during the three years of this study, when stem
22
density increased throughout most strata. An increase in the midstorey during this
23
period is consistent with previous fire experiments in the savannas of northern
24
Australia, Africa and South America, where fire exclusion has enabled the recruitment
20
1
of woody sprouts into the midstorey above the flame zone and away from the
2
competition of grass-layer plants (Brookman-Amissah et al. 1980; San Jose & Farinas
3
1991; Silva et al. 2001; Russell-Smith et al. 2003; Woinarski et al. 2004).
4
A large proportion of individuals at the site, particularly in the grass-layer, were
5
rainforest species; however, it is unclear the extent to which this reflects species
6
composition prior to fire exclusion, or recruitment after. Many of the most common
7
rainforest-associated species (e.g. Alphitonia excelsa and Acacia lamprocarpa) also
8
occur in frequently burnt savannas, but show a strong association with fire protection
9
over time (Bowman & Panton 1995; Woinarski et al. 2004). Edaphic characteristics
10
were critically important in the composition of the midstorey, with substantially higher
11
representation of rainforest species in the South block. These edaphic characteristics
12
clearly influenced the structure and composition of the midstorey within this study site
13
irrespective of fire regime.
14
Grass-layer plant composition has changed during the period of long term fire exclusion
15
since 1981. Vegetation surveys at the time showed that the grass-layer was dominated
16
by the tall annual Sarga intrans (Andersen and Hoffmann 2010), and this species
17
remains dominant immediately outside the Park (Scott 2008). Therefore, in the long-
18
term absence of fire, the composition of the grass understory changed to dominance by
19
the smaller perennial Eriachne triseta and low-growing annuals (e.g.
20
Pseudopogonatherum contortum and Eriachne agrostidea). This change has been
21
documented elsewhere in northern Australia (Russell-Smith et al. 2003; Williams PR et
22
al. 2003a; Woinarski et al. 2004). Mechanistic experiments have independently shown
23
the negative influence of canopy shading and litter cover on the establishment and seed
24
production of grasses, including annual Sorghum (Cook et al. 1998; Scott et al. 2010a).
21
1
Effects of re-introducing fire
2
Reintroducing moderate intensity fire to the long-unburnt savanna caused the
3
vegetation to transition back toward open savanna structure, principally through a
4
reduction in lower midstorey stem density. However, fire intensity was the important
5
factor determining this change. The abundance of midstorey stems increased or
6
remained similar in the UN regime and with repeated, low-intensity fire (Annual Early,
7
mean fire intensity = 116 kW m-1), demonstrating woody plant density had not peaked
8
during the 20 years of fire exclusion. By contrast, the abundance of large saplings
9
decreased in the Late regime (808 and 2121 kW m-1), and the number of lower
10
midstorey stems decreased considerably in both the Biennial Late and Early 05 regime
11
(1110 kW m-1). At least part of this reduction in midstorey stem density could be
12
attributed to top-kill, where individuals remained alive, but were redistributed into
13
smaller size classes (e.g. sprouts). Minimal changes to basal area were recorded (Table
14
2), indicating that most structural change involved individuals with thin stems (i.e. in
15
the midstorey).
16
In general, the resilience of sprouts and small saplings, response of midstorey
17
stems and resistance of overstorey trees to fire in this study is well known in northern
18
Australia. The susceptibility of midstorey stems to burning is consistent with the notion
19
that smaller (lower dbh) plants experience greater rates of stem mortality (top-kill) and
20
have a lower probability of surviving fire (Williams et al. 1999; Werner et al. 2006;
21
Liedloff & Cook 2007; Prior et al. 2009). The fire intensities in this study were low.
22
Had fire intensities been higher (e.g. > 5000 kW m-1), mortality and top-kill may have
23
been much higher and basal area significantly reduced, as a consequence of mortality in
22
1
large overstorey trees, as evidenced by the landscape-scale Kapalga fire experiment
2
(Williams et al. 1999).
3
The reintroduction of fire had a limited effect on grass-layer species
4
composition. Populations of grass-layer plants are highly resilient to fire as a
5
consequence of vegetative regeneration and the annual input of seeds into fire-
6
protected, dry season seed banks (O’Connor 1985; Belsky 1992; O’Connor & Pickett
7
1992; Scott et al. 2010b, Williams PR et al. 2003a). The increase in perennial grass
8
density in compartments remaining unburnt for several years appears to reflect
9
increased seedling recruitment in the absence of fire because there was a uniformly low
10
mortality rate of common, adult perennial grasses recorded at the site in the unburnt
11
regime and the burnt regimes (Scott et al. 2010c). Enhanced seedling recruitment in the
12
unburnt areas could have resulted from an increase in seed production or seed viability,
13
or lower rates of seed or seedling mortality, in those compartments. The marked
14
increase in legume density after fire (Fig. 10 b) was a direct result of elevated
15
temperatures in the soil overcoming physical dormancy (Auld & O’Connell 1991,
16
Morrison et al. 1998, Bell & Williams 1998). The response of seeds to heating has been
17
documented for several legume species in north-eastern Australia, including those at the
18
study site (Scott et al. 2010b; Williams PR et al. 2003b). However, even though
19
perennial grasses and legumes showed significant changes in abundance over time
20
according to fire regime, these life forms were only minor components of the vegetation
21
and so did not significantly influence composition overall.
22
Inter-annual rainfall variability was strongly related to the density of grass-layer
23
plants (Table 4). Rainfall variability has been demonstrated to affect grass-layer species
24
composition in Australia, Africa and South America, particularly in response to long
23
1
periods of drought (O’Connor 1994; O’Connor & Everson 1998; Williams RJ et al.
2
2003; Bisigato & Bertiller 2004). Seedlings of grass-layer plants are vulnerable to
3
desiccation, so mortality would be exacerbated in drier wet seasons with prolonged
4
periods of low rainfall intensity or frequency (Andrew & Mott 1983; O’Connor 1994,
5
1996; Veenendaal et al. 1996). A later start to the wet season, as occurred in 2006-
6
2007, could conversely enhance seedling survival by allowing most seed germination to
7
occur closer to the mid-wet season period of reliable rainfall, thereby minimising post-
8
germination mortality (Crowley & Garnett 1999).
9
10
Reversibility of change following long-term fire exclusion
State-and-transition conceptual models accommodate the possibility that
11
different stable vegetation states are possible within a location, and vegetation state
12
persists unless there is a significant change in an ecosystem driver (Bestelmeyer et al.
13
2003; Briske et al. 2006). An open savanna state is maintained by burning. Fire results
14
in some mortality of midstorey and overstorey stems, and the suppression of woody
15
sprouts in the grass-layer. The open canopy supports a continuous grass-layer, which,
16
due to an annual cycle of fuel accumulation and curing, reinforces the regime of
17
frequent burning (Williams et al. 2003). At the beginning of this study, the savanna
18
structure reflected the effects of a long period of fire exclusion with a well-developed
19
midstorey (2-10m) represented by ~500 – 1000 stems/ha. In our study, fires of
20
moderate intensity (1000 – 2000 kWm-1) reduced the density of the midstorey, with a
21
decrease from 800 to 350 stems/ha in the Biennial Late regime, and from ~1100 to 650
22
stems/ha in the Early 05 regime. This study showed, however, that fire intensity and
23
frequency are critical to these processes, with a maintenance (at ~1000 stems / ha) of
24
the midstorey in the UN regime and increased development of the midstorey (from
24
1
~500 to 750 stems/ha) in the low intensity Annual Early regime. Moderate or high
2
intensity fires over several years may be required to achieve the sparse midstorey
3
characteristic of frequently burnt savanna in this region, and fully restore the structure
4
and / or function of frequently burnt communities typical of the region’s Eucalyptus
5
dominated savanna.
6
The reduction in lower midstorey stems in higher-intensity regimes was driven
7
by the responses of canopy and rainforest species in particular (Table 2). It would
8
seem, therefore, that both canopy and rainforest species experience a recruitment
9
bottleneck in the lower midstorey in these regimes. For the short-term at least, this has
10
little consequence for canopy species because of the resilience and longevity of trees
11
already in the overstorey. Over the longer-term however, fire-free periods will be
12
required for recruitment into the overstorey (Setterfield 2002). Similarly, fire-free
13
periods would be required for rainforest species to dominate the midstorey and
14
overstorey vegetation layers.
15
Even if the flammable annual grass Sorghum intrans was replaced by less
16
flammable grasses such as the perennial Eriachne triseta, the changes still indicate the
17
presence of a pre-threshold state across most of the experimental catchments because
18
the resulting composition carried fires that modified vegetation structure (positive
19
feedbacks). This change in composition and others, such as the increase in perennial
20
grass density with fire exclusion (Fig. 10 a) and increase in legume density after fire
21
(Fig. 10 b), did not cause any structural or functional degradation of the ecosystem. In
22
the time-frame examined, the loss of S. intrans was unidirectional; the reintroduction of
23
fire was unable to reinstate its former dominance. This suggests some longer-term
24
implications, given poor seed dispersal and a lack of long-term seed dormancy within
25
1
the seed bank (Andrew & Mott 1983). The immediate (1 year) post-fire increase in the
2
abundance of Sorghum ecarinatum Lazarides reported on the Arnhem Plateau was
3
probably a result of its occurrence in over half of the quadrats prior to the fire (Russell-
4
Smith et al. 2002).
5
A small proportion of the site (perhaps 10%) appears to have already crossed an
6
ecological threshold, characterised by the predominance of positive feedbacks which
7
reinforce a change between open woodland and closed forest states. There were patches
8
in the South Block (approximately 1000 m2) where woody plant density was very high
9
(~ 10 500 woody stems m-1), presumably a result of more favourable edaphic conditions
10
supporting this midstory development in the absence of fire (Bowman & Minchin 1987;
11
Medina & Silva 1990; Scott et al. 2009). As a consequence, the higher canopy shading
12
and higher amounts of litter (see Fig. 8) resulted in a fuel layer dominated by litter with
13
significantly fewer herbaceous grass-layer plant species (Scott et al. 2009). These litter
14
patches acted as a network of fire breaks, and despite efforts to re-ignite them, were
15
unable to sustain fire. A dominance of litter in the fuel layer therefore reinforced
16
woody thickening on a transition towards closed forest.
17
18
19
Acknowledgements
20
Funding for this project, including a PhD scholarship for K. Scott and a postdoctoral
21
fellowship for C. Parr, was provided by the Bushfire Cooperative Research Centre. A
22
grant from the Bushfire CRC allowed the K. Scott to write this manuscript as a
23
Research Associate at Charles Darwin University. Many people assisted with the
26
1
conduct of experimental fires, vegetation sampling, plant identification and data
2
analysis. We particularly thank Bushfires NT, Jenni Low Choy, Ian Cowie, the late
3
Jack Cusack, Robert Eager, Jon Edgar, Andrew Edwards, Garry Fox, John Henderson,
4
Anthony Kerr, Wawan Sastrawan Manullang, Keith McGuiness, Sally Mitchell, Ken
5
Nicholl, Dave Pearson, Liz Smith and Dick Williams. The Territory Wildlife Park
6
shared our vision towards creating the site as an educational and research facility, and
7
allowed us to access it regularly.
27
1
REFERENCES
2
Andersen, A.N., Cook G.D, Corbett L.K., Douglas M.M., Eager R.W., Russell-Smith J.,
3
Setterfield S.A., Williams R.J., & Woinarski J.C.Z. (2005) Fire frequency and
4
biodiversity conservation in Australian tropical savannas: implications from the
5
Kapalga fire experiment. Austral Ecol. 30, 155–167.
6
Andrew M.H. & Mott J.J. (1983) Annuals with transient seed banks: the population
7
biology of indigenous Sorghum species of tropical north-west Australia. Aust. J.
8
Ecol. 8, 265-276.
9
10
Angassa A. & Oba G. (2008). Effects of management and time on mechanisms of bush
encroachment in southern Ethiopia. African J. Ecol. 46, 186-196.
11
Augustine D.J. & McNaughton S.J. (2004) Regulation of shrub dynamics by native
12
browsing ungulates on East African rangeland. J. App. Ecol. 41, 45-58.
13
14
15
16
Auld T.D. & O’Connell M.A. (1991) Predicting patterns of post-fire germination in 35
eastern Australian Fabaceae. Aust. J. Ecol. 16, 53-70.
Bell D.T. & Williams D.S. (1998) Tolerance of thermal shock in seeds. Aust. J. Bot. 46,
221-233.
17
Belsky A.J. (1992) Effects of grazing, competition, disturbance and fire on species
18
composition and diversity in grassland communities. J. Veg. Sci. 3, 187-200.
19
Bestelmeyer B.T., Brown J.R., Havstad K.M., Alexander R., Chavez G. & Herrick J.E.
20
(2003) Development and Use of State-and-Transition Models for Rangelands. J.
21
Range Mgt. 56, 114-126.
22
23
Bisigato A.J. & Bertiller M.B. (2004) Seedling recruitment of perennial grasses in
degraded areas of Patagonian Monte. Rang. Ecol. Mgt. 57, 191-196.
28
1
2
3
4
5
6
7
8
9
10
11
12
13
Bond W.J. (2008). What limits trees in C4 grasslands and savannas? Ann. Rev. Ecol,
Evol. Syst. 39, 641-659.
Bond W.J. & Parr C.L. (in press) Beyond the forest edge: ecology, diversity and
conservation of the grass biomes. Biol. Cons.
Boone R.B. & Wang G. (2007) Cattle dynamics in African grazing systems under
variable climates. J. Arid Env. 70, 495-513.
Boulière F. (1983) Tropical Savannas. Ecosystems of the World. Volume 13. Elsevier
Publishing, Amsterdam.
Bowman D.M.J.S. & Fensham R.J. (1991) Response of a monsoon-forest savanna
boundary to fire protection, Weipa, northern Australia. Aust. J. Ecol. 16, 111-118.
Bowman D.M.J.S. & Minchin P.R. (1987) Environmental relationships of woody
vegetation patterns in the Australian monsoon tropics. Aust. J. Bot. 35, 151-169.
Bowman D.M.J.S. & Panton W.J. (1993) Factors that control monsoon-rainforest
14
seedling establishment and growth in north Australian Eucalyptus savanna. J.
15
Ecol. 81, 297-304.
16
Bowman D.M.J.S. & Panton W.J. (1995) Munmarlary revisited: Response of a north
17
Australian Eucalyptus tetrodonta savanna protected from fire for 20 years. Aust. J.
18
Ecol. 20, 526-531.
19
20
21
Briske D.D., Fuhlendorf S.D. & Smeins F.E. (2003). Vegetation dynamics on
rangelands: a critique of the current paradigms. J. App. Ecol. 40, 601-614.
Briske D.D., Fuhlendorf S.D. & Smeins F.E. (2006). A Unified Framework for
22
Assessment and Application of Ecological Thresholds. Range. Ecol Mgt. 59, 225-
23
236.
29
1
Brookman-Amissah J., Hall J.B., Swaine M.D. & Attakorah J.Y. (1980) A re-
2
assessment of a fire protection experiment in north-eastern Ghana savanna. J.
3
App. Ecol. 17, 85-99.
4
5
6
Clarke K.R. & Gorley R.N. (2001) Primer v5: User manual/tutorial. Primer-E Ltd.,
Plymouth.
Cook G.D., Ronce O. & Williams R.J. (1998) Suppress thy neighbour: The paradox of a
7
successful grass that is highly flammable, but harmed by fire. In: Proceedings
8
13th Conference on Fire and Forest Meteorology, Lorne, October 1996, pp. 71-4
9
International Association of Wildland Fire, Hot Springs, South Dakota.
10
11
12
Crowley C.M. & Garnett S.T. (1999) Seeds of the annual grasses Schizachyrium spp. as
a food resource for tropical granivorous birds. Aust. J. Ecol. 24, 208-220.
Fensham R.J., Fairfax A.J., Butler D.W. & Bowman D.M.J.S. (2003) Effects of fire and
13
drought in a tropical eucalypt savanna colonized by rain forest. J. Biogeog. 30,
14
1405-1414.
15
16
Holdo R.M. (2007) Elephants, fire and frost can determine community structure and
composition in Kalahari woodlands. Ecol. App. 17, 558-568.
17
Hutley, L.B. and Setterfield, S.A. (2008). Savannas. In Jorgensen, SE and Faith, BD
18
(Ed.), Encyclopedia of Ecology. Oxford, U.K.: Elsevier B.V. (pp. 3143-3154).
19
20
21
Lemmon, P.E. (1956). A spherical densiometer for estimating forest overstory density.
For. Sci. 2: 314-320.
Liddle D.T., Russell-Smith J., Brock J., Leach G.J. & Connors G.T. (1994) Atlas of the
22
vascular rainforest plants of the Northern Territory. Flora of Australia
23
Supplementary Series No. 3. ABRS, Canberra.
30
1
Liedloff A.C. & Cook G.D. (2007) Modelling the effects of rainfall variability and fire
2
on tree populations in an Australian tropical savanna with the FLAMES
3
simulation model. Ecol. Mod. 201, 269-282.
4
Medina E. & Silva J.F. (1990) Savannas of northern South America: a steady state
5
regulated by water-fire interactions on a background of low nutrient availability.
6
J. Biogeog. 17, 403-413.
7
Midgley, J.J., Lawes, M.J. & Chamaille-Jammes, S. (2010). Savanna woody plant
8
dynamics: the role of fire and herbivory, separately and synergystically. Aust. J.
9
Bot. 58, 1-11.
10
Morrison D.A., McClay K., Porter C. & Rish S. (1998) The role of the lens in
11
controlling heat-induced breakdown of testa-imposed dormancy in native
12
Australian legumes. Annals Bot. 82, 35-40.
13
O’Connor T.G. (1985) A synthesis of field experiments concerning the grass-layer in
14
the savanna regions of southern Africa. South African National Scientific
15
Programmes Report 114. CSIR - Commonwealth Scientific and Industrial
16
Research, Pretoria, South Africa.
17
18
O’Connor T.G. (1994) Composition and population responses of an African savanna
grassland to rainfall and grazing. J. App. Ecol. 31, 155-171.
19
O’Connor T.G. (1996) Hierarchical control over seedling recruitment of the bunch-
20
grass Themeda triandra in a semi-arid savanna. J. App. Ecol. 33, 1094-1106.
21
O’Connor T.G. & Everson T.M. (1998) Population dynamics of perennial grasses in
22
African savanna and grassland. In: Population Biology of Grasses (ed G.P.
23
Cheplick) pp. 333-365. Cambridge University Press: Cambridge, UK.
31
1
O’Connor T.G. & Pickett G.A. (1992) The influence of grazing on seed production and
2
seed banks of some African savanna grasslands. J. App. Ecol. 29, 247-260.
3
Parr, C.L. & Andersen, A.N. (2008). Fire resilience of ant assemblages in long unburnt
4
5
savanna of northern Australia. Austral Ecol. 33: 830-838.
Prior L.D., Murphy B.P. & Russell-Smith J. (2009). Environmental and demographic
6
correlates of tree recruitment and mortality in north Australia savannas. Forest
7
Ecol. Mgt. 257, 66-74.
8
9
10
11
Ratnam, J., Bond, W.J., Fensham, R.J., Hoffmann, W.A., Archibold, S. Lehmann,
C.E.R., Anderson, M.T., Higgins, S.I., Sankaran, M. (2011). When is a ‘forest’ a
savanna, and why does it matter? Global Ecol. and Biogeog. 20, 653-660.
Roques K.G., O’Connor T.G. & Watkinson A.R. (2001) Dynamics of shrub
12
encroachment in an African savanna: Relative influences of fire, herbivory,
13
rainfall and density dependence. J. App. Ecol. 38, 268-280.
14
Russell-Smith J., Ryan P.G. & Cheal D.C. (2002) Fire regimes and the conservation of
15
sandstone heath in monsoonal northern Australia: frequency, interval, patchiness.
16
Biol. Cons. 104, 91-106.
17
Russell-Smith, J., Stanton, P.J., Whitehead, P.J. & Edwards, A. (2004) Rain forest
18
invasion of eucalypt-dominated woodland savanna, Iron Range, north-eastern
19
Australia: I. Successional processes. Journal of Biogeography, 31, 1293–1303.
20
Russell-Smith J., Whitehead P.J., Cook G.D. & Hoare J.L. (2003) Response of
21
Eucalyptus-dominated savanna to frequent fires: Lessons from Munmarlary,
22
1973-1996. Ecol. Mon. 73, 349-375.
32
1
2
3
4
5
6
7
8
9
10
11
San José J.J. & Farinas M.R. (1991) Temporal changes in the structure of a
Trachypogon savanna protected for 25 years. Acta Oecol. 12, 237-247.
Scheffer, M. & Carpenter, S.R. (2003) Catastrophic regime shifts in ecosystems:
Linking theory to observation. Trends Ecol. Evol. 18, 648-656.
Scholes R.J. & Archer S.R. (1997) Tree-grass interactions in savannas. Ann. Rev. Ecol.
Syst. 28, 517-544.
Scott K.A. (2008) The distribution and dynamics of grass-layer plants in a tropical
savanna. PhD thesis. Charles Darwin University, Darwin.
Scott K.A., Setterfield S.A., Andersen A.N. & Douglas M.M. (2009) Correlates of grass
species composition in an Australian savanna. Aust. J. Bot. 57, 10-17.
Scott K.A., Setterfield S.A., Douglas M.M. & Andersen A.N. (2010a) Environmental
12
factors influencing the establishment, height and fecundity or the annual grass
13
Sorghum intrans in an Australian tropical savanna. J. Trop. Ecol. 26: 313-322
14
Scott K.A., Setterfield S.A., Douglas M.M. & Andersen A.N. (2010b) Soil seed banks
15
confer resilience to savanna grass-layer plants during seasonal disturbance. Acta
16
Oecol. 36: 202-210
17
Scott K.A., Setterfield S.A., Douglas M.M. & Andersen A.N. (2010c) Fire tolerance of
18
perennial grass tussocks in a savanna woodland. Austral Ecol. 35: 858-861.
19
Setterfield, S.A. (1997) The impact of experimental fire regimes on seed production in
20
21
22
two tropical eucalypt species in northern Australia. Aust. J. Ecol. 22, 279-287.
Setterfield, S. A. (2002). Seedling establishment in an Australian tropical savanna:
effects of seed supply, soil disturbance and fire. J. of Appl. Ecol. 39:949–959.
33
1
Setterfield S.A., Rossiter-Rachor N.A., Hutley L.B., Douglas M.M. & Williams R.J.
2
(2010) Turning up the heat: the impacts of Andropogon gayanus (gamba grass)
3
invasion on fire behaviour in northern Australian savannas. Diversity and
4
Distributions 16, 854–61.
5
Sharp B.R. & Whittaker R.J. (2003) The irreversible cattle-driven formation
6
transformation of a seasonally flooded Australian savanna. J. Biogeog. 30, 783-
7
802.
8
Silva J.F., Zambrano A. & Farinas M.R. (2001) Increase in the woody component of
9
seasonal savannas under different fire regimes in Calabozo, Venezuela. J.
10
11
Biogeog. 28, 977-983.
Sivertsen D.P., McLeod P.J. & Henderson R.L. (1980) Land Resources of the Berry
12
Springs Nature Park and Proposed Development Area. Land Conservation Unit,
13
Northern Territory Government, Darwin.
14
Specht R.L. (1981) Foliage projective cover and standing biomass. Vegetation
15
Classification in Australia (Eds A. N. Gillison & D. J. Anderson), pp. 10–21.
16
CSIRO Publishing, Canberra.
17
18
19
Stringham T.K., Krueger W.C. & Shaver P.L. (2003) State and transition modelling: An
ecological process approach. J. Range Mgt. 56, 106-113.
Veenendaal E.M., Ernst W.H.O. & Modise G.S. (1996) Effect of seasonal rainfall
20
pattern on seedling emergence and establishment of grasses in a savanna in south-
21
eastern Botswana. J. Arid Env. 32, 305-317.
34
1
Vigilante T. & Bowman D.M.J.S. (2004) Effects of fire history on the structure and
2
floristic composition of woody vegetation around Kalumburu, North Kimberley,
3
Australia: a landscape-scale natural experiment. Aust. J. Bot. 42, 381-404.
4
Walker B.H., Langridge J.L. & McFarlane F. (1997) Resilience of an Australian
5
savanna grassland to selective and non-selective perturbations. Aust. J. Ecol. 22,
6
125-135.
7
Werner P.A. (2005) Impact of feral water buffalo and fire on growth and survival of
8
mature savanna trees: An experimental field study in Kakadu National Park,
9
northern Australia. Austral Ecol. 30, 625-647.
10
Werner P.A., Cowie I.D. & Cusack J.S. (2006) Juvenile tree growth and demography in
11
response to feral water buffalo in savannas of northern Australia: An experimental
12
field study in Kakadu National Park. Aust. J. Bot. 54, 1-14.
13
14
15
Westoby M., Walker B. & Noy-Meir I. (1989) Opportunistic management for
rangelands not at equilibrium. J. Range Mgt. 42, 266-274.
Williams P.R., Congdon R.A., Grice A.C. & Clarke P.J. (2003a) Effect of fire regime
16
on plant abundance in a tropical eucalypt savanna of north-eastern Australia.
17
Austral Ecol. 28, 327-338.
18
Williams P.R., Congdon R.A., Grice A.C. & Clarke P.J. (2003b) Fire-related cues break
19
seed dormancy of six legumes of tropical eucalypt savannas in north-eastern
20
Australia. Austral Ecol. 28, 507-514.
21
Williams R.J., Cook G.D., Gill A.M. & Moore P.H.R. (1999) Fire regime, fire intensity
22
and tree survival in a tropical savanna in northern Australia. Aust. J. Ecol. 24, 50-
23
59.
35
1
Williams R.J., Müller W.J., Wahren C., Setterfield S.A. & Cusack J. (2003) Vegetation.
2
In: Fire in tropical savannas: the Kapalga experiment (Eds. A.N. Andersen, G.D.
3
Cook & R.J. Williams) pp.79-106. Springer, New York.
4
Woinarski J.C.Z., Risler J. & Kean L. (2004) Response of vegetation and vertebrate
5
fauna to 23 years of fire exclusion in a tropical Eucalyptus open forest, Northern
6
Territory, Australia. Austral Ecol. 29, 156-176
36
Table 1. Burning schedule for experimental fire regimes at the Territory Wildlife
Park. Values are mean intensity (kW m-1) with 1 SE in parenthesis (data
unavailable for Early 04). – indicates no fire event.
Fire regime
Year
2004
2005
2006
Annual Early
144.0 (18.1)
77.7 (36.7)
125.3 (54.1)
Early 04
n/a
-
-
Early 05
-
1110.3 (569.0)
-
Biennial Late
808.4 (625.0)
-
2121.3 (975.3)
37
Table 2. Summary ANOVA results for variation in species richness, and the abundance of woody plants in each stratum between 2004 and
2007. Basal area, species richness and standing dead stems consider all strata together, whereas abundance variables are categorised by stratum
for all species combined (e.g. sprouts, 0-0.49 m), and then by species groups based on the maximum height that they grow too (i.e. groundlayer, shrub, tree species) within each stratum. Abundance was also analysed for rainforest species). Significant F-ratios are indicated by
asterisks (*P < 0.05, **P < 0.01, ***P < 0.001). Grass-layer species in the small sapling stratum, overstorey species in the large sapling
stratum, and monsoon forest species in the overstorey stratum, were too uncommon for separate analysis.
Response variable
F-ratios
Year (Y)
Fire regime (R)
Block (B)
Y×R
Y×B
F (1, 10)
F (4, 10)
F (10, 15)
F (4, 10)
F (10, 15)
Species richness
12.98**
1.49
2.24
2.37
1.46
Standing dead stems
23.91***
0.66
4.53**
1.27
2.41
11.28**
0.76
5.19**
1.58
1.56
10.52**
0.73
8.13***
1.81
1.14
1.10
0.31
1.67
2.14
0.44
Shrub
10.62**
0.73
3.41*
1.34
2.40
Tree
9.31*
0.74
4.24**
4.98*
0.07
5.39*
1.04
12.35***
1.24
3.76*
Live stem abundance
Grass-layer
Sprouts (0.0-0.49 m)
Rainforest
Ground-layer
Small saplings (0.5-1.99
38
m)
Rainforest
1.25
0.04
11.84***
1.58
0.55
Shrub
3.68
0.76
11.16***
1.22
2.76*
19.27**
1.82
2.62*
2.41
0.50
0.32
0.54
1.96
3.89*
1.70
0.49
0.45
4.57**
2.31
1.08
< 0.01
0.42
2.01
3.42
1.46
7.07*
3.44
2.13
5.19*
5.42**
Rainforest
5.27*
0.44
6.32***
1.30
7.96***
Shrub
4.78
3.56*
1.80
4.45*
3.47*
Tree
17.09**
1.45
2.68*
12.64***
0.37
2.46
0.37
3.38*
2.18
3.44*
Rainforest
0.75
0.19
7.03***
1.77
3.37*
Shrub
1.03
0.42
2.01
1.86
5.62**
Tree
4.39
1.29
1.13
1.37
0.54
2.50
0.79
3.58*
2.91
1.42
Tree
Midstorey
Large saplings (> 2.0
m & < 3 cm dbh)
Rainforest
Shrub
Lower Midstorey
(2.0-4.99 m)
Upper Midstorey
(5.0-9.99 m)
Overstorey (10.0 m +)
39
Table 3. Summary for the repeated measures ANOVA analysis of grass-layer species richness, total density, and the density of individual life
form groups. Statistically significant test statistics are indicated by asterisks (*P < 0.05, **P < 0.01, ***P < 0.001). †Data examined
annual surveys between 2004 and 2007 inclusive; all other variables examined data between 2005 and 2007 inclusive; late sampling in
2004 precluded the sampling of some life form group.
Response variable
Source of variation and F-value
Fire regime (R)
Block (B)
Year (Y)
R×B
R×Y
B×Y
R×B×Y
F (4, 8)
F (2, 15)
F (2, 4)
F (8, 15)
F (8, 16)
F (4, 30)
F (16, 30)
2.4
74.0***
4.0
5.9**
2.1
7.9***
3.0**
2.0
18.1***
16.1*
1.9
1.2
4.0*
2.2*
Annual grasses
0.9
28.3***
27.9**
4.1**
1.8
6.4***
3.6**
Perennial grasses†
2.2
77.9***
F (3, 6) = 4.1
4.4**
F(12,24) = 2.2*
F(6,45) = 3.2*
F24,45=1.5
Forbs
2.3
27.8***
1.1
2.5
2.2
2.6
0.7
Sedges
1.7
18.3***
5.7
1.9
0.9
2.6
2.4*
Legumes
3.7
34.9***
8.0*
4.1**
3.9**
4.8**
3.7**
Species richness
Abundance
Total
40
Table 4. Inter-annual variation in the abundance of grass-layer plants (categorised by life
form; fire regimes pooled) and rainfall in the late dry season (August to October inclusive)
and wet season (November to February inclusive). Abundance data for most life-form groups
are unavailable for 2004, as many annual species had already cured by the time of sampling.
Year
2004
2005
2006
2007
Total
-
60.1 (7.0)
62.2 (7.0)
86.7 (11.7)
Annual grasses
-
16.0 (3.0)
23.8 (4.7)
38.3 (6.0)
7.0 (0.6)
7.7 (0.6)
9.1 (0.8)
11.3 (1.1)
Forbs
-
17.6 (2.9)
18.5 (3.0)
24.1 (7.0)
Sedges
-
15.8 (2.8)
6.8 (1.4)
8.5 (1.9)
Legumes
-
3.0 (0.5)
3.9 (0.7)
5.4 (0.7)
64
101
33
2
1050
581
789
1115
Late dry season prior
7
6
5
1
Wet season prior
80
62
67
78
Abundance (individuals m-2 ± 1 SE)
Perennial grasses
Rainfall (mm)
Late dry season prior†
Wet season prior‡
Rainfall (no. rain days)
†Rainfall collected within 2 km of the study site. ‡Rainfall at the site was not consistently recorded in
one wet season (2007) and so data presented are from Middle Point AWS 36 km to the north-east
(Bureau of Meteorology, unpublished data). Mean rainfall at Middle Point AWS (for the period 19572004) is 71 mm on 9 rain days in the late dry season period, and 965 mm on 72 rain days for the wet
season period (Bureau of Meteorology, unpubl. Data)
41
1
Table 5. Mean density (untransformed) of woody and grass-layer plant species in 2007, in each block at the Territory Wildlife Park study site, which
2
contributed at least 3% of the dissimilarity in species composition between blocks. Species in bold font were most abundant in the ecotone community
3
(Southern block). Note separate SIMPER analyses were conducted for woody plants and grass-layer plants. † Monsoon forest species (Liddle et al. 1999)
Species
Structural / lifeform
Mean density by block
Dissimilarity
Cumulative
group
(stems ha-1, woody plants;
(%)
dissimilarity
individuals ha-1, grass-layer plants)
North
Centre
South
(woodland)
(woodland)
(ecotone)
(%)
Woody plants
Exocarpus latifolius†
Shrub
319.4
4,586.8
13,506.9
10.05
10.05
Petalostigma pubescens
Shrub
125.0
17.4
3,208.3
5.82
15.87
Calytrix exstipulata
Shrub
1,975.7
2,885.4
2,006.9
4.93
20.8
Alphitonia excelsa†
Shrub
152.8
159.7
2,534.7
4.77
25.57
Petalostigma quadriloculare
Shrub
1,034.7
1,590.3
1,423.6
3.76
29.33
Syzygium eucalyptoides†
Tree
194.4
27.8
1,197.9
3.09
32.43
42
Grass-layer plants
Pseudopogonatherum
Annual grass
28,2824.1
15,9861.1
0.0
contortum
Thaumastochloa major
Eriachne triseta
Schizachyrium fragile
Fimbristylis sp. Charles
10.5
Annual grass
9,8611.1
64,907.4
3,379.6
5.64
16.15
Perennial grass
20,277.8
5,7731.5
139,675.9
5.45
21.6
Annual grass
98,518.5
6,296.3
3,796.3
5.18
26.77
Sedge
65,972.2
60,000.0
1,574.1
3.72
30.49
3.65
34.14
Darwin (JL Egan 5300)
Spermacoce exserta
10.5
Herb
9,351.9
106,064.8
79,398.2
43
1
Figure captions
2
Fig. 1. The Territory Wildlife Park study site with savanna with a Eucalyptus dominated
3
overstory and a (a) moderate to (b) well-developed midstorey.
4
Fig. 2. Total abundance of woody stems within each height stratum (sprouts, < 0.5 m;
5
small saplings, 0.5-1.99 m; large saplings, > 2 m but < 3 cm dbh; lower midstorey, 2.0-
6
4.99 m; upper midstorey, 5.0-9.99 m; overstorey, 10.0+ m), in 2004 before regime
7
treatments were first implemented (white bars), and three years later in 2007 (black
8
bars). Data are presented by block: a) North (woodland), b) Centre (woodland), and c)
9
South (ecotone). Data are mean stem density ± SE, where transect (n = 10) is the unit of
10
replication. Note differences in scale.
11
Fig. 3. Total abundance of woody stems of monsoon forest species (sensu Liddle et al.
12
1994) within each height stratum (sprouts, < 0.5 m; small saplings, 0.5-1.99 m; large
13
saplings, > 2 m but < 3 cm dbh; lower midstorey, 2.0-4.99 m; upper midstorey, 5.0-9.99
14
m), in 2004 before regime treatments were first implemented (white bars), and three
15
years later in 2007 (black bars). Data are presented by block: a) North (woodland), b)
16
Centre (woodland), and c) South (ecotone). Data are mean stem density ± SE, where
17
transect (n = 10) is the unit of replication. Rainforest species were very uncommon in
18
the overstorey stratum and data are note presented. Note differences in scale.
19
Fig. 4. Mean (+ 1 SE) values of a) litter cover and b) canopy cover, during the four
20
years (2004-2007) of the manipulative fire experiment. Arrows and associated numbers
21
indicate the timing of fires for each treatment and their intensity, and ‘n/a’ indicates that
22
fire intensity was not assessed (kW m-1 ± 1 SE; see Table 1).
23
Fig. 5. Total abundance of stems within each height stratum above the sprout layer (small
24
saplings, 0.5-1.99 m; large saplings, > 2 m but < 3 cm dbh; lower midstorey, 2.0-4.99
25
m; upper midstorey, 5.0-9.99 m; overstorey, 10.0+ m), in 2004 before regime treatments
26
were first implemented (black bars), and three years later in 2007 (white bars). Data are
27
presented by fire regime treatment: a) Annual Early, b) Early 05, c) Biennial Late, d)
28
Early 04, and e) UN (see Table 1 for details of intensity and frequency). Data are mean
29
stem density ± SE, where transect (n = 6) is the unit of replication. Note differences in
30
scale.
44
1
Fig. 6. Total abundance of sprouts (0-0.49 m) within each of the fire regime treatments
2
(see Table 1 for details of intensity and frequency) of the experiment, in 2004 before
3
regimes were first implemented (white bars), and three years later in 2007 (black bars).
4
Fig. 7. Species richness of all woody species within each of the fire regime treatments
5
(see Table 1 for details of intensity and frequency) of the experiment, in 2004 before
6
regimes were first implemented (white bars), and three years later in 2007 (black bars).
7
Fig. 8. Total abundance of woody stems of monsoon forest species (sensu Liddle et al.
8
1994) within each height stratum (sprouts, < 0.5 m; small saplings, 0.5-1.99 m; large
9
saplings, > 2 m but < 3 cm dbh; lower midstorey, 2.0-4.99 m; upper midstorey, 5.0-9.99
10
m; overstorey, 10.0+ m), in 2004 before regime treatments were first implemented
11
(white bars), and three years later in 2007 (black bars). Data are presented by fire
12
regime treatment: a) Annual Early, b) Early 05, c) Biennial Late, d) Early 04, and e) UN
13
(see Table 1 for details of intensity and frequency). Data are mean stem density ± SE,
14
where transect (n = 6) is the unit of replication. Note differences in scale.
15
Fig. 9. The abundance of basal resprouts in 2007 as a proportion of live stems within
16
each of the fire regime treatments (Annual Early, Early 05, Biennial Late, Early 04 and
17
UN).
18
Fig. 10. Mean (+ 1 SE) density of (a) perennial grasses, (b) legumes, and (c) Sorghum
19
intrans, categorised by fire treatment and year. Perennial grasses and legumes showed a
20
significant regime × year interaction in the experiment (see Table 1), whereas S. intrans
21
did not. Arrows and associated numbers indicate the timing of fires for each treatment
22
and their intensity (kW m-1 ± 1 SE; R. Williams, unpublished data; see Table 1).
45
Figure 1
a)
b)
46
Figure 2
a) Sprouts
North
Centre
South
Centre
South
b) Small saplings
North
c) Large saplings
North
Centre
South
47
d) Lower midstorey
North
Centre
South
e) Upper midstorey
North
Centre
South
Centre
South
f) Overstorey
North
48
Figure 3
a) Sprouts
North
Centre
South
b) Small saplings
North
c) Large saplings
Centre
South
49
North
Centre
South
d) Lower midstorey
North
Centre
South
e) Upper midstorey
North
Centre
South
50
Figure 4
a)
100
90
125 ± 54
78 ± 37
80
70
60
1110 ± 569
144 ± 18
2121 ± 975
808 ± 625
50
40
30
20
Litter cover (%)
10
0
04
05
06
07
04
05
Annual early
06
07
04
Early 05
05
06
07
Biennial late
100
90
80
70
n/a
60
50
40
30
20
10
0
04
05
06
07
04
05
Early 04
b)
85
06
07
Unburnt
144 ± 18
1110 ± 569
78 ± 37
80
125 ± 54
75
808 ± 625
70
2121 ± 975
65
60
55
50
Canopy cover (%)
45
40
04
05
06
07
04
05
Annual early
06
07
Early 05
80
n/a
70
65
60
55
50
45
40
04
05
Early 04
06
07
04
05
06
Unburnt
05
06
Biennial late
85
75
04
07
07
51
Figure 5
a) Annual Early
b) Early 05
52
c) Biennial Late
d) Early 04
e) UN
53
Figure 6
54
Figure 7
55
Figure 8
a) Sprouts
b) Small saplings
c) Large saplings
56
d) Lower midstorey
e) Upper midstorey
57
Figure 9
25
Proportion of live stems (%)
20
15
10
5
0
Annual Early
Early 05
Biennial Late
Early 04
Unburnt
58
Figure 10
20
125 ± 54
Perennial grass density (individuals m
-2
)
a)
15
78 ± 37
144 ± 18
1110 ± 569
808 ± 625
2121 ± 975
04
06
10
5
04
05
06
07
04
Annual Early
05
06
07
Early 05
Biennial Late
20
15
10
5
04
05
Early 04
06
07
04
05
05
06
Unburnt
07
07
59
b)
14
2121 ± 975
12
10
125 ± 54
808 ± 625
Legume density (individuals m -2)
8 144 ± 18
78 ± 37
6
1110 ± 569
4
2
0
05
06
07
05
06
Annual Early
07
05
06
Early 05
07
Biennial Late
14
12
10
8
6
4
2
0
05
06
07
05
06
Early 04
c)
07
Unburnt
0.6
0.5
Sorghum intrans density (individuals m -2)
0.4
0.3
0.2
144 ± 18
125 ± 54
78 ± 37
1110 ± 569
808 ± 625 2121 ± 975
0.1
0.0
04
05
06
07
04
Annual Early
05
06
07
Early 05
0.5
0.4
0.3
0.2
0.1
0.0
05
06
Early 04
07
04
05
06
Unburnt
05
06
Biennial Late
0.6
04
04
07
07