Response of rock-outcrop and fringing vegetation to disturbance by fire and drought

Page 1

CSIRO PUBLISHING

www.publish.csiro.au/journals/ajb Australian Journal of Botany, 2007, 55, 736–748

Response of rock-outcrop and fringing vegetation to disturbanceby fire and drought

Andrew Benwell

PO Box 641, Mullumbimby, NSW 2482, Australia. Email: [email protected]

Abstract. A comparative study of vegetation responses to fire and drought investigated whether species regenerationmode, seedling density response (SDR) and seedling shoot growth rate varied significantly in relation to disturbance agent(fire and drought), habitat type (rock-outcrop and fringing vegetation) and plant growth form. A three-way ANOVA showedthat SDR varied significantly in relation to all three categorical variables and most strongly in relation to disturbance agent.Seeders comprised 87% of the post-fire flora and 99.3% of the peak seedling population in rock-outcrop habitat, whileresprouters were much more prominent in fringing woodland. Species SDRs and seedling growth rates were generallymuch higher after fire. Fire produced a high SDR with high shrub, grass and ephemeral therophyte components, anddrought produced a low SDR dominated by grasses and herbs and inhibited shrub and ephemeral recruitment. Post-fireobligate-seeder shrubs behaved as facultative resprouters after drought. Some species exhibited SDRs equivalent to fireand drought, others appeared to require fire for regeneration, while others recruited more successfully after drought. Thisspectrum of responses indicated a range of optimal disturbance environments, depending on species, which was onlypartly consistent with the hypothesis that species exhibit essentially the same life-history syndromes in response to fireand drought. The dominance of seeders in outcrop vegetation appeared to be related to skeletal soil, higher disturbancefrequency and soil trophic conditions, rather than low fire frequency.

Introduction

Rock outcrops represent a unique environment for plant survival,characterised by very shallow soil overlying bedrock. In non-arid regions of Australia, rock outcrops generally support plantcommunities that are floristically quite distinct from surroundingvegetation (Hopper et al. 1997; Hunter 1999). In some regions,particularly in central eastern and south-western Australia, asubstantial number of plant species are endemic to rock-outcropvegetation, apparently owing to the selection pressures operatingin this localised and divergent habitat (Hopper et al. 1997;Benwell 2004).

Physical agents of disturbance such as fire, drought and storm,which cause punctuated episodes of widespread damage to aboveground vegetation, are a key element of the terrestrial plantenvironment. These disturbance events regulate and influencea wide range of plant-community processes including mortality,recruitment, secondary succession, biogeochemical cycling andcompetition (Bormann and Likens 1979; White 1979; Sousa1984; Huston 1994). In Australia, drought and fire are reportedto be the main environmental disturbances having an impact onrock-outcrop vegetation (Hopper et al. 1997; Main 1997), just asin the wider landscape, although not necessarily with the samefrequency and intensity.

Studies of fire response in rock-outcrop vegetation ongranite geology in central eastern and south-western Australiahave found that compared with surrounding vegetation, thesecommunities have a high proportion of species with seederlife-history strategies and few resprouters (Hopper et al. 1997;Hunter 1999; Clarke and Knox 2002). Seeders and resprouters

represent broadly different types of fire response or life-historysyndrome; resprouters survive fire and regenerate vegetatively,with or without seedlings, and seeders are killed by fire andregenerate from seed (Gill 1981; Bell 2002). Considerable life-history diversity exists within these two general categories;e.g. seeders include therophytes and long-lived, stress-tolerantseeders (Benwell 2004). Insight into the factors regulating thedivision of plant communities into seeders and resprouters,or life-history groups, is of practical value in vegetationmanagement (Keith 1996) and also relevant to the wider studyof how plant communities are assembled and species diversitymaintained (Huston 1994).

Clarke and Knox (2002) postulated that divergence in thelife-history composition of plant communities may be related todifferences in fire frequency and/or regeneration niche amonghabitats. The fire-frequency model proposes that the dominanceof rock-outcrop communities by seeder species is linked withlower fire frequency and assumes that seeder species have beenreduced in the surrounding landscape by higher fire frequencyand increased in rock outcrops by relative fire protection(Hopper 2000; Clarke 2002; Hunter 2003). This hypothesis isbased on the following two premises: first, that fire frequency issignificantly less in rock outcrop habitat because the mosaic ofvegetation and exposed rock present a barrier to the spread offire, and second, that seeders require relatively low fire frequencyto persist.

The second type of model proposed by Clarke and Knox(2002) to explain the different proportions of seeders andresprouters in different habitats revolves around the concept of

© CSIRO 2007 10.1071/BT07050 0067-1924/07/070736

Page 2

Response of rock-outcrop and fringing vegetation to disturbance Australian Journal of Botany 737

regeneration niche (Grubb 1977). Four aspects of regenerationniche, determined by type of habitat, were considered ofpotential importance in regulating proportions of seeders andresprouters, including soil nutrients, soil texture and moisture,openness and rainfall variability. Clarke and Knox (2002)concluded that none of these mechanisms could explaindifferences in fire response across all habitats and favoured acombination of disturbance-frequency and regeneration-nichemodels.

Hunter (2003) linked the predominance of seeders in rock-outcrop vegetation to harsh environment, low disturbancefrequency and high light levels. Grime (1993) considered that thesignificance of different plant life-history strategies (e.g. seedersand resprouters) generally relates to patterns of resourceavailability in the edaphic environment (i.e. regeneration niche)and species strategies of resource capture and utilisation. Toexamine differences in the soil nutrient component of theregeneration niche that may influence proportions of seeders andresprouters in rock-outcrop and fringing vegetation, the presentstudy included soil nutrient analyses of both habitats during post-fire regeneration.

A further aim of the present study was to evaluate whetherspecies and community responses to fire and drought were thesame or significantly different. Plant fire responses such asbradysporous seed release, mass germination of soil seedbanks,resprouting from protected buds and mast-flowering are usuallyviewed as fire-adaptive responses, as they appear to promote thesurvival, reproduction and recruitment of species in fire-pronevegetation (Gill 1981; Gill and Groves 1981; Gill and Bradstock1992). However, agents of disturbance such as drought and stormblow-down elicit qualitatively similar responses (Gill 1981;Florence 1996), which suggests that specific life-history traitsoften viewed as fire-adaptive may have multiple evolutionarycauses including any other disturbance that results in the deathof above-ground plant parts. If plant life-history traits exhibitedin response to fire and drought are essentially the same, it ispossible that dry adaptations may have pre-adapted elementsof Australia’s flora to increases in fire frequency that occurredin the late Tertiary and Quaternary (Kershaw 1985; Singh andGeissler 1985; Atahan et al. 2004). Significant differences inspecies response to different disturbance agents would alsohave implications for models seeking to explain the proportionsof seeders and resprouters in plant communities (Clarkeand Knox 2002), as species persistence may be affected bydisturbance frequency rather than specifically fire frequency. Inthe case of rock-outcrop vegetation, which is more desiccation-prone because of shallow soil, the frequency of disturbancemay be determined by the combined impact of drought andfire events.

There is a large amount of information on the vegetationdynamics and life-history traits of Australian plants in responseto fire, but relatively little on response to drought. In semi-arid woodlands of south-western Western Australia, Yates et al.(1994) found that landscape-scale disturbances such as droughtwere important regulators of recruitment and stand structure.Hnatiuk and Hopkins (1980) found that drought in kwonganshrubland caused widespread mortality and that both vegetativeregeneration and seedling regeneration were effective meansof surviving the disturbance. Overall, they noted the similarity

between the response of shrubland to drought and fire. George(2002) described drought-survival strategies and the browningand regreening responses (‘diallagy’) of vegetation in south-western Western Australia, to the very dry season of 2000–2001,the first time this response had been recorded across a wide areaof the landscape.

Opportunities to examine the impacts of both fire anddrought on vegetation at the same location seldom arise becauseof the lengthy time frames involved and unpredictability ofdisturbance events. Comparison of fire and drought responsesat the same location was made possible in the present studyby the occurrence of a drought 2 years after fire, duringwhat began as a fire-response monitoring program (Benwell2004). Species and community responses to fire and droughtwere compared in terms of life-history traits exhibited in thepost-disturbance regeneration phase. These included speciesmode of regeneration, seedling density response and seedlingshoot growth rate. The aim of the study was to measure theresponses of rock-outcrop and fringing vegetation to fire anddrought and to utilise these data to explore the following tworesearch questions:

(1) are patterns of seeder and resprouter composition in thestudy area consistent with other regions and what are theecological processes driving the patterns observed; and

(2) are species responses to fire and drought the same orfundamentally different?

Materials and methodsStudy areaThe study area is located at Bald Knob Mountain, 90 kmnorth-west of Lismore in the New South Wales North CoastBioregion. Bald Knob (altitude 828 m) is composed of ashallow intrusion of rhyolite, an acid volcanic rock of similarmineralogy to granite, but finer in grain size (Blatt and Tracey1996). Responses to fire and drought were studied in pavementshrubland and fringing woodland habitats on the northern sideof the mountain at 600–700-m altitude. Pavement shrublandcomprises a mosaic of bare rock and shrubland growing onskeletal soil ∼5–30 cm deep and reaching a height of 2–3 mat maturity. Shrubland changes abruptly to woodland aroundthe margin of the outcropping rock mass where increasingsoil depth supports trees from 4 to 15 m high (e.g. Eucalyptusnotabilis, E. campanulata, E. eugenoides, Corymbia gummiferaand Allocasuarina littoralis) and an understorey of shrubs,grasses and herbs. Beyond this ecotone, which is relativelynarrow, the vegetation grades into tall open forest. Meanannual rainfall is 1200 mm, two-thirds falling in summerand autumn and one-third in winter and spring (Bureauof Meteorology at www.bom.gov.au).

Disturbance eventsFire responses were monitored following a hazard reductionburn carried out by State Forests of New South Wales inOctober 1998. Drought responses were monitored after a periodof below-average rainfall between autumn and late spring 2000,which caused widespread desiccation and dieback of pavementshrubland. The fire burnt about half the total area of shrublandand much of the surrounding forest. Fire intensity was estimated

Page 3

738 Australian Journal of Botany A. Benwell

from flame scorch height and the degree of incineration of shootsystems as moderate in both pavement shrubland and fringingwoodland. Time since fire in rock outcrop and fringing habitatwas estimated from vegetation maturity (e.g. bradysporous seedstorage) to be 5–10 years.

Rainfall in the early regeneration phase after fire and droughtwas similar. After both fire and drought, the dry spring periodended in late October and more than 150 mm fell by the end ofDecember. Rainfall in the first 5 months after fire was 568 mm(October–February) and 498 mm during the correspondingperiod after drought. Average rainfall for this period is 558 mm(Bureau of Meteorology, Woodenbong Station).

Life-history traitsSpecies responses to fire and drought were recorded in termsof regeneration mode, seedling density response and seedlingshoot growth rate, as defined below.

Regeneration modeSpecies regeneration mode was classified in terms of primary

regeneration response and source of seedling regeneration.Three primary responses were recognised after Naveh (1975)—obligate seeder (regrowth by seedlings only), facultativeresprouter (regrowth by seedlings and reshooting) and obligateresprouter (regrowth by reshooting only). Sources of seedlingregeneration were the soil seedbank, canopy seedbank orpost-disturbance flowering (Gill 1981). Primary regenerationresponse was determined from stands subject to 100% crownscorch or combustion by fire, and 100% drought induceddefoliation (of normally evergreen plants).

Seedling density response (SDR)This trait was defined as the approximate maximum seedling

density exhibited by species after disturbance and was recorded5 months after fire and drought. After fire, total communityseedling density and the seedling density of most species peakedat ∼5 months (Fig. 1). Since fire and drought occurred inthe same season (late spring) and post-disturbance rainfall

Mea

n no

. of s

eedl

ings

m–2

Months after fire

Fig. 1. Flux in mean total seedling density during the first 29 months afterfire in rock-outcrop shrubland and fringing woodland. The column on theright-hand side indicates seedling recruitment after the drought in year 2000.Bars indicate s.e.

was similar, seedling density response after drought was alsomeasured at 5 months.

Seedling shoot growth rateThe seedling shoot growth rate of shrub and tree species was

measured in terms of maximum seedling height per quadrat.

Monitoring fire responseLife-history traits were monitored in 1-m2 quadrats by astratified, random sampling design. Five 20 × 20 m plots wereplaced in burnt pavement shrubland and three in burnt fringingwoodland. Two 1-m2 quadrats were then positioned randomlyin each plot quarter with pairs of random numbers between0 and 9 (m). With eight fixed quadrats in each plot, a total of40 quadrats were monitored in rock-outcrop habitat and 24 infringing woodland (Table 1). Life-history traits were recorded 2,5, 10, 23 and 29 months after fire between December 1998and March 2001. At 2 months, only total seedling numberwas recorded, as the recently emerged seedlings were initiallydifficult to identify. By 5 months, species were identifiable fromthe first true leaves as well as cotyledon shape, size and colour. At23 months, small numbers of seedlings recruited in the previousrainy season, evident by size and the presence of cotyledons,were counted and excluded from the first-year cohort. The censusat 23 months was carried out at the start of September, beforethe most intense period of the 2000 drought in the latter partof September and early October. Maximum seedling heights ofspecies per quadrat were recorded at each monitoring. Mode ofregeneration was determined by excavating plants outside plots.

Monitoring drought responseFive months after the drought ended, five 20 × 20 m plots wereplaced in patches of unburnt pavement shrubland missed by theOctober 1998 fire, which showed pronounced, drought-induceddieback. Three plots were placed in unburnt fringing woodlandadjacent to shrubland plots. Eight fixed 1-m2 quadrats wereestablished in each plot by the same stratified random methodand life-history traits recorded as for post-fire monitoring. Asecond census was carried out in September 2002 to assess ratesof seedling establishment.

Table 1. Types of disturbance monitored and number of quadrats(1 m2) used to sample different habitats and disturbances during

the monitoring program

Disturbance Types of disturbancemonitored

No drought Drought(pre-2000 drought) (post-2000 drought)

Fire (burnt Oct. 1998) + +No Fire (unburnt Oct. 1998) – +

Habitat Number of quadratsFire Drought Drought + fire

Pavement shrubland 40 40 40Fringing woodland 24 24 24

Page 4

Response of rock-outcrop and fringing vegetation to disturbance Australian Journal of Botany 739

Several microhabitat factors potentially affecting post-drought SDR at the scale of quadrats were recorded to model thefactors influencing post-drought seedling density in pavementshrubland, including (i) total number of plants and number ofdead plants per shrub species, (ii) extant species crown cover,(iii) pre-drought species crown cover, (iv) total (overlapping)vegetation crown cover, (v) percentage exposed rock and(vi) percentage dense litter. Individuals of all shrub species werecounted in each quadrat and classed as either ‘dead’, ‘100%defoliated and reshooting’ or ‘partly defoliated/no evidenceof defoliation’. The low number of species in rock-outcropshrubland made it possible to identify species from remainingdead leaves, bark type and growth form with a reasonabledegree of accuracy. Small stems (<30 cm tall) were omitted fromcounts because of increased likelihood of error in identification,with the exception of the low-growing species Leucopogonneo-anglicus. Extant and pre-drought species crown cover andtotal (overlapping) crown cover (Walker and Hopkins 1990)were estimated visually. Pre-drought species crown cover wasestimated as the combined cover of surviving plants and deadframes.

Monitoring drought response soon after fireThe response to drought of recently burnt vegetation wasrecorded in the fire response plots 5 months after drought(29 months after fire). Post-fire and post-drought seedlingcohorts were easily distinguished by seedling size and presenceof cotyledons. As previously, all seedlings were identified andcounted and maximum species heights per quadrat measured. Infringing woodland, species were identified to growth form only(e.g. grasses, shrubs and so on).

Disturbance frequencyRainfall and temperature records for Woodenbong station(10 km east of Bald Knob) spanning 68 years were acquiredfrom the Bureau of Meteorology to estimate the average returnperiod of droughts as or more intense than the 2000 drought. Soilmoisture availability in rock-outcrop vegetation was assumedto be governed mainly by recent rainfall (as there is minimalsoil water storage) and mean temperature, the latter varyinglittle from year-to-year in comparison with precipitation. Totalrainfall for August and September, the driest months of the yearon average, was used as a crude drought intensity index. Averagerainfall for the August–September period is 76.3 mm.

Soil nutrient analysisSoil samples were collected from the monitoring plots inpavement shrubland and fringing woodland. The same stratifiedrandom sampling method was followed except that one samplewas collected from each plot quarter and an additional plot wasestablished in fringing woodland, yielding a total of 20 samplesfrom pavement shrubland and 16 from fringing woodland.A 5 × 5 × 8 cm column of topsoil (surface litter excluded)was collected at each sampling point and placed in a sealed,labelled plastic bag. The soil samples were collected 2.5 yearsafter fire (April 2001). Soil depth in pavement shrubland wasmeasured at four random points per quadrat by driving in

a steel spike and measuring depth to bedrock. A soil augerwas used to take samples at 20- and 50-cm depths in fringingwoodland and to determine soil depth to a maximum of 80 cm(limit of auger).

Samples were air-dried, sieved and 100-g samples weighedout for analysis. Analyses determined pH, % organicmatter, electrical conductivity (µS cm−1), total N (mg g−1),total P (mg g−1), available P (mg g−1), nitrate (mg g−1),ammonia (mg g−1), Ca (mg g−1), K (mg g−1), Mg (mg g−1) andNa (mg g−1), following the methods by Gale and Hoare (1991).

Data analysisComparison of community regeneration mode spectraTo examine whether frequencies of regeneration mode in

pavement shrubland and fringing woodland were significantlydifferent, a two-way classification of species by primaryregeneration mode and habitat was tested with chi-square.

SDR in relation to disturbance, habitat and growth formRelationships between ‘seedling density response’ and the

categorical variables ‘disturbance agent’ (fire, drought andfire + drought), ‘habitat’ (pavement shrubland and fringingwoodland) and ‘plant growth form’ (trees, shrubs, grasses,perennial herbs, ephemerals/annuals and all species) wereanalysed by three-way ANOVA. The analysis was based onthe general linear model and a Model I design in which allfactors were fixed-effect factors (Hays 1981). Main effectsand all two- and three-way interactions were analysed foreach categorical variable. Interactions occur when the effect ofone input variable is influenced by the level of another inputvariable. The response variable was log-transformed to reduceheteroscadicity and a plot of residuals against fitted y-valueswas inspected to confirm that the analysis met the ANOVAassumption of equality of variance (Moore and McCabe 1999).Mean SDRs for different disturbance agents, habitat types andgrowth forms were calculated.

Microhabitat factors influencing post-droughtseedling densityMicrohabitat variables potentially affecting post-drought

seedling density at the scale of quadrats were modelled bysimple and stepwise, multiple linear regression (Moore andMcCabe 1999). The response variable ‘mean seedling density’(untransformed and log) was regressed against the followingfactors: (i) shrub mortality rate, or proportion of individualskilled by drought; (ii) crown cover, or extant overlapping percentcrown cover; (iii) crown cover change defined as 1 minus(sum of species extant crown covers/sum of species pre-droughtcrown covers); (iv) percentage cover of exposed rock; and(v) percentage cover of dense litter.

Stepwise multiple regression selects variables in order ofreduction of sums of squares (variances) in the linear model.Since stepwise multiple regression can distort the relativeimportance of independent variables when they are correlated,simple linear regressions were also applied to determine the‘total relationship’ of a given independent variable with thedependent variable (Tabachnick and Fidell 1989).

Page 5

740 Australian Journal of Botany A. Benwell

Soil nutrient analysisSoil nutrient attributes in pavement shrubland and fringing

woodland habitat were analysed for significant differences byone-way ANOVA based on the general linear model.

Results

Vegetation response to fire

Mode of regeneration

The number of seeder species was about the same inrock-outcrop and fringing vegetation, but the number ofresprouter species was much greater in fringing woodland(χ2 = 6.069, d.f. = 1, P = 0.014, Table 2). Only four speciesof resprouter were present in rock-outcrop habitat and onlytwo of these were common (Leptospermum microcarpum andLepidosperma clipeicola), whereas fringing woodland had20 resprouter species. No obligate resprouters were recorded.Two species exhibited ecotypic variation in regenerationmode (Entolasia stricta and Keraudrenia hillii), behaving asseeders in pavement shrubland and as resprouters in fringingwoodland. The seeder component in pavement shrublandincluded several ephemeral therophytes, including Monotaxismacrophylla, Bulbostylis densa, Muehlenbeckia gracillima,Actinotus gibbonsii, Polymeria calycina and Cyperus sp. ‘BaldKnob’. This life-history group was absent from fringingvegetation.

Seedling recruitment

Mean total SDR was higher in pavement shrublandthan fringing woodland (Fig. 1), but not significantly sountil 10 months (t = 3.21, d.f. = 61, P = 0.002). In pavementshrubland, seeders accounted for 99.3% of the total SDRand resprouters only 0.7% (Fig. 2). In fringing woodland,resprouters accounted for 64.5% of the SDR and seeders 35.5%.The bradysporous resprouter Leptospermum polygalifoliumsubsp. cismontanum contributed 45.1% of the SDR in fringingwoodland.

Shrubs were the most abundant growth-form component ofthe total seedling population in pavement shrubland, followedby grasses, ephemerals, perennial herbs and trees. Shrubswere also most abundant in fringing woodland, followed bygrasses, herbs and trees. The most abundant species in pavementshrubland was the grass Entolasia stricta, followed by shrubsin the Casuarinaceae, Myrtaceae, Faboideae, Sterculiaceae andMimosoideae. Herbs in the Euphorbiaceae, Cyperaceae andApiaceae were also common. These families also accounted formost seedlings in fringing woodland but species were generallydifferent (Fig. 2).

Seedling shoot growth rate

Seeders grew more rapidly than resprouters in both habitats(Fig. 3). Fringing-woodland species that germinated in rock-outcrop habitat after fire, grew well initially and then died outduring the 2000 drought (Fig. 3). Pooling seedling height datafor all woody species (tall shrubs and trees), mean maximumseedling height was 35% higher in pavement shrublandthan fringing woodland after 5 months (30.8 cm ± 1.4 s.e. v.18.8± 1.7) and 300% higher after 23 months (68.6 cm ± 2.6 v.23.9 ± 1.6). This result reflects the higher number of resprouterspecies in fringing woodland and the slower shoot growth rateof resprouter seedlings.

Response to drought

Mortality and mode of regeneration

The mortality rate of shrub species in mature pavementshrubland during drought averaged 56.3 ± 5.4%. Speciesmortality rates ranged from 100% in Allocasuarina rigida to12.1% in Leptospermum microcarpum (Fig. 4). Five months afterdrought, mean summed shrub cover was 56.5 ± 4.9% less thanthe estimated pre-drought value. Shrub species that responded asseeders after fire behaved as facultative resprouters after drought.For example, of the 82% of stems of Leptospermum variable(a post-fire obligate seeder) that were defoliated by drought,approximately half re-shot and this response was accompaniedby a SDR of 7.55 seedlings m−2 (Table 5). Little dieback occurredin fringing woodland where vegetation was buffered againstwater stress by greater soil depth.

Seedling recruitment

The total post-drought SDR in pavement shrubland wasabout one-third the total post-fire SDR (Fig. 5). In fringingwoodland, post-drought SDR was much less, only 1.6% ofthe post-fire response. Post-drought seedling recruitment inpavement shrubland was dominated by grasses, with muchreduced shrub and ephemeral components (Fig. 5). A fewspecies of grass and herb exhibited enhanced SDR afterdrought, e.g. Digitaria breviglumis and Trachymene incisa(Table 5). Mature Allocasuarina rigida, which experienced100% mortality (in plots), released large quantities ofbradysporous seed after drought but the SDR was very low(Table 5). Nevertheless, 3 years after fire, fire-recruited seedlingshad thinned out from approximately 40 to five times the post-drought cohort at 2 years (∼5 v. 1 seedling m−2), reducing thediscrepancy in seedling density. Leptospermum variable alsoreleased large quantities of bradysporous seed and exhibited asimilar low SDR compared with after fire (Table 5).

Table 2. Spectrum of species regeneration responses in pavement shrubland and fringing woodland at Bald Knob,defined by primary strategy and seed source

Habitat Obligate seeder Facultative resprouter Total speciesSoil Canopy Soil Canopy Post-fire

seedbank seedbank seedbank seedbank flowering

Pavement shrubland 23 3 3 1 0 30Fringing woodland 28 3 13 4 3 51

Page 6

Response of rock-outcrop and fringing vegetation to disturbance Australian Journal of Botany 741

See

dlin

gs m

–2

Plant species

(a)

(b)

Fig. 2. Species exhibiting the 15 highest seedling densities 5 and 23 monthsafter fire in (a) pavement shrubland and (b) fringing woodland. Standard errorbars refer to 5-month data. OSR, obligate seed regenerator; FR, facultativeresprouter.

Seedling shoot growth rate

In pavement shrubland, the seedling shoot growth rates ofcommon shrub species were significantly less after drought.Mean maximum seedling height 23 months after drought was∼50% less than for the same period after fire in Leptospermumvariable, L. microcarpum and Jacksonia sp. Little Plain and 70%less in Allocasuarina rigida (Table 3).

Response to drought shortly after fire

Mode of regeneration

The 2000 drought, which occurred 2 years after fire,accelerated mortality of the post-fire seedling cohort in pavementshrubland, which declined by 75% between 23 and 29 months(Fig. 1). Post-fire seeders behaved as resprouters after droughtand in this case the reshooting response was exhibited by 1.5–2-year-old seedlings rather than adult plants. Many seedlingsre-shot after being defoliated, for example, Keraudrenia hillii(69.2% of the surviving cohort), Jacksonia sp. ‘Little Plain’(30.5%), Leptospermum variable (25.32%), Acacia brunioides(24.8%) and Allocasuarina rigida (8.0%).

Plant species

Mea

n m

ax. s

eedi

ng h

eigh

t m–2

(a)

(b)

Fig. 3. Mean maximum seedling heights of the fastest-growing woodyspecies in (a) pavement shrubland and (b) fringing woodland rankedaccording to height at 23 months (bars indicate s.e.). Species without anincrement for 29 months had either decreased in mean maximum height ordied out (d). OSR, obligate seed regenerator; FR, facultative resprouter.

Plant species

Per

cent

age

Fig. 4. Drought responses of common shrub species in pavementshrubland. Individuals were classed as unaffected (white bars, foliageretained), defoliated (hatching, foliage shed then re-shot) or dead(black bars, foliage shed, no recovery observed). Allocasuarina rigida,Acacia brunioides, Jacksonia sp. ‘Little Plain’, Leptospermum variable,Leucopogon neo-anglica and Leptospermum microcarpum.

Page 7

742 Australian Journal of Botany A. Benwell

Mea

n no

. of

seed

lings

m–2

after fire after drought after fire + drought

after fire + droughtafter droughtafter fire

160

140

120

100

80

60

40

20

0

160

140

120

100

80

60

40

20

0

160

140

120

100

80

60

40

20

0

160

140

120

100

80

60

40

20

0

160

140

120

100

80

60

40

20

0

160

140

120

100

80

60

40

20

0

(a) (b) (c)

(d) (e) (f )

Pavement shrubland

Fringing woodland

Growth-form category

Fig. 5. Mean seedling density in pavement shrubland and fringing woodland after fire, drought and fire + drought. Total seedling density response (SDR,±s.e.) is shown in the top right. T, tree; S, shrub; H, perennial herb; G, grass; E, ephemeral; trace (<1.0 seedlings per square metre).

Table 3. Mean maximum seedling height (cm ± s.e.) of four common shrub species, 5 and 23 months afterdrought and fire

Type of Time since Jacksonia sp. Leptospermum Leptospermum Allocasuarinadisturbance disturbance ‘Little Plain’ variable microcarpum rigida

(months)

Fire 5 29.16 ± 2.26 16.27 ± 1.93 6.91 ± 1.46 21.89 ± 1.87Drought 5 18.93 ± 3.09 10.42 ± 0.90 6.27 ± 0.43 18.75 ± 2.78Fire 23 64.67 ± 4.17 39.20 ± 2.82 29.33 ± 8.78 78.43 ± 4.13Drought 23 29.80 ± 3.25 18.54 ± 3.57 13.38 ± 1.70 23.00 ± 2.63

Seedling recruitment

The total SDR in recently burnt shrubland affected by droughtwas significantly less than that after drought or fire in matureshrubland (Fig. 5). The grass Digitaria breviglumis accountedfor 50% of the seedling population in pavement shrubland. Infringing woodland, the SDR was very small and similar to thedrought (without fire) response.

Total SDR in relation to major study variables

Three-way ANOVA showed that the main effects ofdisturbance type, habitat and growth form on SDR were allhighly significant (P < 0.000), as were all two- and three-way interaction terms (Table 4). The F-statistic in Table 4indicates that differences in SDR were greatest with respectto disturbance agent, followed by growth form then habitat.The main effects of disturbance agent, habitat type and growthform on SDR were described above and are shown graphicallyin Fig. 5.

Species SDR in relation to disturbance type

The majority of species exhibited higher SDRs after fire,although there were several exceptions, e.g. the grass Digitaria

Table 4. Three-way ANOVA results for mean seedling density responseagainst disturbance agent, growth form and habitat type

Main factors and interactions are given under ‘source’. The F-ratios indicateof the relative statistical significance of each term. ***P < 0.001

Source F-ratio Significance

Disturbance agent 351.57 ***Growth form 127.13 ***Habitat type 50.96 ***Disturbance × growth form 17.92 ***Disturbance × habitat 20.43 ***Growth form × habitat 28.88 ***Disturbance × growth 13.41 ***

form × habitat

breviglumis and the herb Trachymene incise, which hadsignificantly higher SDRs after drought (Table 5). Of the threestructurally dominant species, Leptospermum microcarpum(a resprouter) exhibited a higher SDR after drought, but thedifference was not significant, and Leptospermum variableand Allocasuarina rigida recruited much more prolificallyafter fire. Several other species, including Pimelea linifolia,

Page 8

Response of rock-outcrop and fringing vegetation to disturbance Australian Journal of Botany 743

Table 5. Species seedling density responses (±s.e.) in pavement shrubland after fire, drought and drought + fire, sorted accordingto post-fire SDR

Growth form: G, grass; S, shrub; E, ephemeral; H, perennial herb; T, tree

Species Growth Regeneration mode Fire Drought Drought + fireform post-fire

Entolasia stricta G OSR 82.65 ± 15.41 15.88 ± 3.37 11.13 ± 3.50Allocasuarina rigida S OSR 46.85 ± 17.31 1.30 ± 1.10 0Leptospermum variable S OSR 32.69 ± 8.62 7.55 ± 2.82 0.03 ± 0.03Jacksonia sp.‘Little Plain’ S OSR 26.25 ± 5.94 2.68 ± 0.80 0.10 ± 0.08Keraudrenia hillii S OSR 23.10 ± 6.66 0.05 ± 0.04 0.15 ± 0.07Acacia brunioides S OSR 20.88 ± 3.63 0.15 ± 0.06 0.03 ± 0.03Digitaria breviglumisA G OSR 14.15 ± 5.79 43.70 ± 10.89 25.63 ± 9.14Monotaxis macrophylla E OSR 13.53 ± 5.30 0.38 ± 0.12 0Cyperus sp. ‘Bald Knob’ E OSR 7.18 ± 5.75 0.25 ± 0.25 0Pultenaea villosa S OSR 3.98 ± 1.82 0 0Goodenia paniculata H OSR 3.78 ± 1.71 0 0Actinotus gibbonsiiA E OSR 2.13 ± 1.10 1.55 ± 1.38 8.95 ± 4.57Trachymene incisaA H OSR 2.00 ± 1.74 21.38 ± 7.05 2.50 ± 1.65Gonocarpus teucroides S OSR 1.30 ± 0.61 0 0Leptospermum microcarpumB S FR 0.57 ± 0.28 1.00 ± 0.32 0Dodonaea viscosa subsp. spatulata S OSR 0.88 ± 0.54 0 0Bulbostylis densa E OSR 0.83 ± 0.46 0.23 ± 0.20 0Hovea longifolia S OSR 0.80 ± 0.48 0 0Polymeria calycina E OSR 0.75 ± 0.33 0.03 ± 0.03 0Lepidosperma urophorum H FR 0.65 ± 0.24 0.03 ± 0.03 0.13 ± 0.09Acacia viscidula S OSR 0.63 ± 0.36 0 0Allocasuarina littoralis T OSR 0.48 ± 0.27 0 0Eucalyptus notabilis T FR 0.48 ± 0.18 0 0Dodonaea boronifolia S OSR 0.20 ± 0.08 0 0Leucopogon neo-anglicusB S FR 0.10 ± 0.05 0.60 ± 0.19 0Aristida jerichoensis G OSR 0.05 ± 0.05 0.13 ± 0.13 0Hypochaeris radicata X OSR 0.05 ± 0.05 0.03 ± 0.03 0.03 ± 0.03Pimelea linifolia S OSR 0.05 ± 0.03 0.88 ± 0.71 0Crassocephalum crepidoides X OSR 0.03 ± 0.02 0.03 ± 0.03 0.05 ± 0.04Eragrostis browniiA G OSR 0.03 ± 0.02 1.03 ± 0.63 0Micrantheum ericoides S OSR 0.03 ± 0.02 0 0Ozothamnus diosmifolium S OSR 0.03 ± 0.02 0 0

ASpecies that exhibited an enhanced response after drought or drought + fire.BData 23 months post-disturbance.

Leucopogon neo-anglicus, Eragrostis brownii, Aristidajerichoensis and Actinotus gibbonsii, exhibited higherSDRs after drought or drought soon after fire (Table 5).However, differences were generally non-significant, reflectingpatchy distribution, low frequency of occurrence and highstandard errors.

Post-drought seedling density response in relationto microhabitat variables

The simple linear regression model with the highest explainedvariance was log(seedling density) on ‘stems dead’ (Table 6,r2 (adj.) = 54.0%). This indicated an exponential relationshipbetween seedling recruitment density and drought intensityof the form y = aebx , where y equals seedling density, e is alogarithmic base and x the degree of drought-induced stemmortality (Fig. 6). Stepwise multiple regression found thatthe predictor variables explained 71.1% of the variance inlog(seedling density), with ‘stems dead’ again selected as themost influential variable, followed by litter cover and total crowncover.

Table 6. Results of simple linear regression of mean seedling densityagainst five predictor variables

Response % Variance explained∼predictor [r2 (adjusted)]

Mean seedling density ∼stems dead 29.8Log(mean seedling density) ∼stems dead 54.0∼Summed cover change (decrease) 43.3∼Total crown cover 41.0∼Exposed rock 4.9∼Litter 7.7

Drought frequency

In terms of August–September rainfall, the frequency ofdrought as or more intense than the 2000 event is about onceevery 12 years on average (Fig. 7). High-intensity drought,defined as less than 10-mm rainfall in August–September,occurred twice in the 68-year record for Woodenbong (1991and 1941).

Page 9

744 Australian Journal of Botany A. Benwell

Stems dead

Log

(see

dlin

g de

nsity

)

3

2

1

0

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Fig. 6. Fitted-line plot and 95% confidence interval of the relationshipbetween proportion of stems killed by drought and log(seedling density).

20001990198019701960195019401930

200

100

0

Year

Rai

nfal

l Aug

–Sep

t (m

m)

Mean76.3 (43.6)

Drought

Fig. 7. Drought index based on August–September rainfall forWoodenbong station. Reference lines indicate mean August–Septemberrainfall (s.d. in parentheses) and August–September rainfall during thedrought in year 2000.

Soil nutrient analysis

Pavement shrubland soil had significantly higher total N,total P, available P, NO3, NH4, organic matter and electricalconductivity than fringing vegetation soil (Table 8). Available Pwas almost five times higher in pavement shrubland and nitratewas eight times higher (Table 7). There was no significantdifference in Ca and K, whereas Mg and Na were significantlyhigher in fringing soil. Soil pH was significantly more acidin pavement shrubland (4.72 ± s.e. 0.06 v. 5.40 ± 0.05). Thesmall number of samples taken from subsoil in fringing soilat Bald Knob indicated declining total N, total P, available Pand ammonia with depth down to 50 cm, and other nutrientsshow illuviation at 20 cm and eluviation by 50 cm (Table 8).Both pavement shrubland and fringing woodland were burnt2.5 years before collection of soil samples. Lower concentrationsof nutrients in fringing soil may reflect more intense competitionfor nutrients by the much greater plant biomass and nutrientadsorption by the much greater volume of clay dominated solum.Higher nutrient levels in pavement shrubland could be dueto high soil organic matter, seepage from fringing woodlandupslope of outcrops and concentration of nutrients because

of restricted soil volume and less loss of nutrient to verticalleaching.

Discussion

Variation in seeder and resprouter composition amonghabitats

The seeder and resprouter composition of rock outcrop andfringing vegetation at Bald Knob, an acid volcanic residual,was similar to outcrop vegetation on granite in other regionsof Australia (Hopper et al. 1997; Clarke and Knox 2002) andon other continents (Wyatt and Fowler 1977; Porembski andBarthlott 2000). Seeders made up 87% of the flora in pavementshrubland after fire, and the fringing woodland ecotone hadsignificantly more resprouters (39%). Other research indicatesthat 65–85% of species in open forests of the surroundinglandscape behave as resprouters in response to medium-to high-intensity fire (Christensen and Kimber 1975; Warket al. 1987; Benwell 2000; Clarke and Knox 2002). Togetherthese findings indicate a gradient in life-history compositionrelated to soil depth, whereby seeders increase in proportionas soil depth decreases, while resprouters follow the oppositetrend, increasing as a proportion of the community as soildepth and volume increase. The habitat factor most stronglyinfluenced by soil depth is presumably soil moisture availability(Bratton 1976).

Apart from the obvious factor of soil depth and moistureavailability, soil nutrient analysis demonstrated other significantdifferences in the edaphic environment of rock outcrop andfringing vegetation, which would give support to the theorythat differences in the ‘regeneration niche’ may be drivingdifferences in the proportions of seeders and resproutersamong habitats. Although the soil nutrient analysis providedonly a snapshot view of nutrient dynamics within the solum,differences in the major plant nutrients N and P were verymarked 2.5 years after fire, with available P in the topsoil almostfive times higher in pavement shrubland and nitrate eight timeshigher. These soil nutrient conditions would be particularlysuited to seeders, which have a life-history characterised by rapidgrowth and high reproductive output (Pate 1993; Bell 2002), asevidenced by the seedling shoot growth rate and SDR data fromthis study.

Evaluation of the low-fire frequency model (Clarke andKnox 2002) is limited by lack of detailed measurements offire regimes in rock-outcrop and surrounding vegetation. Highfire frequencies have been reported from some rock-outcroplocalities (e.g. Hunter 1999; Hopper 2000) and the twoconsecutive fires studied by Yates et al. (2003) were describedas extensive and burnt much of the rock-outcrop vegetation,which would appear to contradict the idea that exposed rockacts as a fire barrier. It seems reasonable to assume that acontinuous expanse of exposed rock would prevent the spreadof low-intensity fire, but it may not prevent at least partial spreadof medium- to high-intensity fire because of ‘fire spotting’or burning embers falling ahead of the fire front (Florence1996). Yates et al. (2003) reported marked differences in theabundance of obligate seeders after the two fires but this wasattributed to protracted drought after one of the fires, rather thanfire frequency.

Page 10

Response of rock-outcrop and fringing vegetation to disturbance Australian Journal of Botany 745

Tab

le7.

Res

ults

ofon

e-w

ayA

NO

VA(G

LM

)fo

rdi

ffer

ence

sin

12so

ilnu

trie

ntat

trib

utes

intw

oso

ilty

pes

(pav

emen

tlit

hoso

land

outc

rop

frin

ging

soil)

atB

ald

Kno

b**

*P<

0.00

1;**

0.00

1<

P<

0.01

,*0.

01<

P<

0.05

;n.s

.,no

tsig

nific

ant

Fact

orN

o.of

pHE

CTo

talN

Tota

lPA

vaila

ble

PN

O3

NH

4%

Org

anic

Ca

KM

gN

asa

mpl

es(µ

Scm

−1)

(mg

g−1)

(mg

g−1)

(mg

g−1)

(mg

g−1)

(mg

g−1)

mat

ter

(mg

g−1)

(mg

g−1)

(mg

g−1)

(mg

g−1)

Soil

type

3664

.77*

*1.

38n.

s.29

.35*

**22

.01*

**21

.33*

**3.

32n.

s.11

.22*

*10

.66*

*0.

18n.

s.3.

04n.

s.30

.70*

**18

.05*

**(p

avem

ent

v.fr

ingi

ng)

Tab

le8.

Mea

ns(±

s.e.

)fo

r12

soil

nutr

ient

attr

ibut

esin

diff

eren

tso

ilty

pes

Soil

type

/N

o.of

pHE

CTo

talN

Tota

lPA

vaila

ble

PN

O3

NH

4%

Org

anic

Ca

KM

gN

aSo

illo

catio

nsa

mpl

es(µ

Scm

−1)

(mg

g−1)

(mg

g−1)

(mg

g−1)

(mg

g−1)

(mg

g−1)

mat

ter

(mg

g−1)

(mg

g−1)

(mg

g−1)

(mg

g−1)

dept

h(c

m)

Pave

men

t20

4.72

141.

2285

14.0

426

6.63

5.89

9.66

46.2

328

.55

84.6

411

5.50

36.9

243

.08

14.7

(0.0

6)(7

8.28

)(7

17.7

5)(1

4.47

)(0

.88)

(4.0

2)(4

.56)

(3.5

6)(2

2.84

)(8

.54)

(5.5

7)(2

.50)

(0.6

3)Fr

ingi

ng16

5.40

38.1

536

48.9

516

0.35

1.28

1.40

28.3

514

.85

71.9

514

5.09

92.9

563

.60

80+

(0.0

5)(2

.43)

(448

.03)

(17.

75)

(0.1

9)(0

.56)

(1.7

5)(1

.43)

(16.

91)

(15.

73)

(8.9

2)(4

.41)

A2

hori

zon

45.

4123

.43

1121

.41

91.1

00.

481.

4419

.78

7.20

1.61

87.4

987

.19

46.0

2–

(20

cm)

(0.0

8)(0

.60)

(241

.70)

(13.

68)

(0.1

1)(0

.69)

(3.6

4)(0

.76)

(0.9

3)(1

0.17

)(8

.78)

(6.4

5)B

1ho

rizo

n4

5.32

28.4

036

8.01

60.0

30.

511.

1111

.14

6.05

2.77

126.

9728

8.30

60.1

5–

(50

cm)

(0.1

1)(7

.74)

(34.

86)

(5.4

2)(0

.19)

(0.0

9)(1

.07)

(1.0

7)(3

.26)

(34.

51)

(31.

32)

(4.4

3)

Page 11

746 Australian Journal of Botany A. Benwell

Also implicit in the fire-frequency model is the notion thatseeders require low fire frequency to persist at a locality, ‘low’meaning inter-fire intervals greater than the maturation periodof the seeders. Excluding ephemerals, the great majority ofobligate seeders in fire-prone vegetation in Australia reachmaturity in 5 years and very few require more than 8 years(e.g. Benson 1985; Benwell 1998). Similar or shorter maturationperiods (e.g. fire ephemeral species) were observed during thepresent study, indicating that seeders generally do not needparticularly low fire frequency, certainly not of the order ofdecades. Seeders most at risk from short-interval fires are slowlymaturing species with canopy seedbanks (e.g. Proteaceae andCasuarinaceae) rather than soil-stored seed (Keith 1996; Woolleret al. 2002). The slowest-maturing seeder in pavement shrublandat Bald Knob was Allocasuarina rigida, which was estimatedto require 8–10 years for formation of a population replacingbradysporous seedbank, similar to its congener in fringingwoodland (A. littoralis). The critical fire interval that can causesignificant changes in the abundance of the majority of (non-ephemeral) seeder species appears to be less than 5 years ormore than 15 years (Gary and Morrison 1995). The low-firefrequency model would need to demonstrate that fire frequencyin surrounding forest was on average less than 5 years (forseeders to concentrate on rock outcrops). However, for muchof the range country of eastern New South Wales, average firefrequency in open forest is 10 years or more (Williams and Gill1995). As well as the potential threat to seeders from high firefrequency, it is noted that long fire intervals can also disadvantageobligate seeders because of seedbank decay (Auld 1987) andby exceeding the longevity of species with bradysporous seedstorage. Finally, the low-fire frequency model also implies thatrock outcrops are subject to low disturbance frequency (Hunter2003); however, this model overlooks the impact of drought,which together with fire is likely to produce a relatively highfrequency of disturbance. Overall, there appears to be littleevidence in terms of either actual low-fire frequency data foroutcrop habitat or seeder requirement for low fire frequency(e.g. >1 in 10 years) in order to persist at a site, to support thelow-fire frequency hypothesis for the high proportion of seedersin outcrop habitat.

The predominance of seeders in rock-outcrop habitat impliesthat the life-history attributes that distinguish seeders fromresprouters, such as their survival mechanism, growth rateand reproductive effort, have a greater selective advantage inrock-outcrop habitat. This is reinforced by the two speciesrecorded at Bald Knob with infraspecific seeder and resprouterecotypes, where the seeder population was associated with rockoutcrop and the resprouter with fringing woodland. Seeder andresprouter seedlings exhibit highly divergent patterns of growthand dry-matter partitioning in the first years of growth (Pate1993). For example, Hansen et al. (1991) compared seederand resprouter species of Bossiaea and found that 6-year-oldseedlings of a seeder species were 18 times greater in size (dryweight) than were those of the resprouter species. The essentialfeature distinguishing seeder and resprouter life histories appearsto be the strategy of resource capture and utilisation. Seedersexploit temporary conditions of high resource availability causedby disturbance and survive periods of resource scarcity and

intense competition by stress avoidance, often as dormant seed.Resprouters are adapted to survive under conditions of resourcescarcity and intense competition (Grime 1979). The resprouteris disadvantaged in terms of seedling shoot growth and survival;however, established plants can recover from fire by using storedphotosynthate. In rock-outcrop habitat, perennating organs suchas rhizomes may succumb to desiccation in shallow outcrop soilso that seed becomes the most viable survival strategy.

The predominance of seeders in rock-outcrop vegetationappears to be a consequence of skeletal soil, unstableedaphic environment (particularly extreme fluctuations in soilmoisture), frequent disturbance and loose biogeochemicalcycling. Dormant seed becomes a more effective survivalstrategy than vegetative persistence in this harsh environment.Paradoxically, this combination of habitat factors operating in anon-arid environment also creates temporary trophic conditionsafter disturbance that favour species with high rates of seedlingestablishment, rapid growth and early maturation, or obligate-seeder life histories. A diversity of resprouter life historiesappears to require a relatively stable edaphic environment,which is lacking in rock-outcrop habitat because of shallowsoil and extreme fluctuations in available soil moisture. Theprimary control of seeder and resprouter composition in rock-outcrop and adjoining habitats is considered to lie in patternsof resource availability, particularly soil-nutrient dynamics, andspecies strategies of resource capture and utilisation in differenthabitats (Grime 1993; Pate 1993) rather than fire frequency.Fire frequency can influence seeder and resprouter abundancein ecological time (Gary and Morrison 1995; Keith 1996);however, the primary process broadly inferred above, maydetermine seeder–resprouter composition in different habitatsacross evolutionary time.

Response to fire and drought disturbance

The impacts of fire and drought were comparable only in rock-outcrop habitat, as the impact of below-average rainfall in 2000on fringing woodland did not represent a disturbance in thesense of causing widespread plant mortality. The response ofpavement shrubland to fire and drought was markedly different.Fire produced a much higher SDR, dominated by shrubsand ephemerals. Drought produced a weak SDR dominatedby grasses and with markedly lower shrub and ephemeralcomponents. Drought disturbance acting on the early stageof post-fire regeneration stimulated recruitment in only partof the herbaceous flora and elicited virtually no response inthe woody flora. Drought disturbance accelerated the thinningof seedling populations and with the exception of a fewrapidly maturing grasses and herbs, inhibited recruitment. Adisturbance regime where drought was the predominant agentwould presumably produce a rock-outcrop community similarto the grass-dominated, granite-outcrop communities describedby Hunter and Clarke (1998) for the north-western slopes regionof New South Wales.

The lower SDR recorded after drought may reflect theintensity of disturbance rather than fundamental differences inspecies responses to agents of disturbance. The 2000 droughtthat killed an average of 56% of shrubs in pavement shrubland

Page 12

Response of rock-outcrop and fringing vegetation to disturbance Australian Journal of Botany 747

at Bald Knob (in plots) was a less intense disturbance thanfire and appears to have been a relatively low-intensity droughtevent according to climatic records. The positive exponentialrelationship detected by regression analysis between SDR anddrought-induced stem mortality indicates that a much higherSDR would eventuate after severe drought events that occurabout twice in a century.

The regeneration mode of seeder species was distinctlydifferent after fire and drought disturbances. Fire caused almostcomplete mortality of populations, and seeder species respondedwith high levels of seedling recruitment. After drought, most‘seeders’ behaved as resprouters, with low levels of seedlingrecruitment. Again, differences in mode of regeneration mayreflect disturbance intensity. The impact of the 2000 drought onplant tissues may have been equivalent to a very low-intensityfire. Species behaving as obligate seeders after medium- tohigh-intensity fire often resprout after very low-intensity fire(Gill 1981; Benson 1985). In a severe drought, more plantsare likely to be killed by desiccation, with less recovery byresprouting. The lower seedling-growth rate of common shrubspecies recorded after drought indicates significant differencein soil trophic conditions after drought and fire, and suggests asudden release of nutrients after fire and a more gradual releaseafter drought.

As the drought and fire plots at Bald Knob were located atdifferent points (unlike fire and drought after fire plots), it ispossible that ‘significant’ differences in response to fire anddrought were due to sampling patchy distributions. However,this is unlikely for common species. For common species itwas found that some exhibited a depressed response to drought(Jacksonia sp. ‘Little Plain’ and Leptospermum variable),others exhibited almost total inhibition of seedling recruitmentafter drought (Keraudrenia hillii and Acacia brunioides),others recruited significantly greater numbers after drought(Trachymene incisa and Digitaria breviflora) and some appearedto recruit equally well after fire or drought (Leptospermummicrocarpum).

Lowest SDRs were recorded for drought soon after fire.This would have been partly due to depletion of seedbanks bythe fire response; however, the soil seedbank must have beensubstantially replenished by ephemerals, grasses and perennialssuch as Keraudrenia hillii, which flowered prolifically in the first6 months after fire. Of the three common ephemeral therophytes,Monotaxis macrophylla, Cyperus sp. ‘Bald Knob’ and Actinotusgibbonsii, the first two exhibited very low SDRs after droughtand a complete absence of response after drought + fire,suggesting specific adaptation to the post-fire environment. Thethird ephemeral, Actinotus gibbonsii, recruited equally well afterfire and drought and exhibited a higher (non-significant) SDRafter drought + fire. Since the fire plots were the same as thedrought + fire plots, seedbanks of all three ephemeral speciesmust have been present in the soil seedbank before fire anddrought + fire disturbances.

Overall, variation in SDR indicated different optimaldisturbance environments for seedling recruitment, dependingon species. Several species appeared to have evolved specificfire-requiring adaptations, some species exhibited equivalentresponses to fire and drought, and others recruited moresuccessfully after drought disturbance. Species primarily

responsive to fire may have evolved adaptations to exploit soilnutrient enrichment in the early post-fire, edaphic environment(Siddiqui et al. 1976; Attiwill and Leeper 1987), which maydiffer from drought where nutrients are likely to be releasedmore slowly from dead plant material by organic decomposition.Interpretation of the data was confounded by the apparentlyunequal disturbance intensity of the fire and drought eventscompared. The situation is a complex one, only partly consistentwith the hypothesis that species exhibit essentially the same life-history syndromes in response to fire and drought.

References

Atahan P, Dodson JR, Itzstein-Davey F (2004) A fine-resolution Pliocenepollen and charcoal record from Yallalie, south-western Australia.Journal of Biogeography 31, 199–205.

Attiwill PM, Leeper GW (1987) ‘Forest soils and nutrient cycles.’(Melbourne University Press: Melbourne)

Auld TD (1987) Population dynamics of the shrub Acacia suaveolens (Sm)Wild: survivorship throughout the life cycle, a synthesis. AustralianJournal of Ecology 12, 139–151.

Bell DT (2002) Ecological response syndromes in the flora of SouthwesternWestern Australia: fire resprouters versus reseeders. Botanical Review67, 417–441.

Benson DH (1985) Maturation periods for fire sensitive shrub species inHawkesbury sandstone vegetation. Cunninghamia 1, 339–349.

Benwell AS (1998) Post-fire seedling recruitment in coastal heathlandin relation to regeneration strategy and habitat. Australian Journalof Botany 46, 75–101. doi: 10.1071/BT94046

Benwell AS (2000) ‘Vegetation survey of Nymboida National Park.’ (NSWNational Parks and Wildlife Service: Glenn Innes, NSW)

Benwell AS (2004) Maintenance of diversity in volcanic outcrop vegetationof the Moreton Region Central Eastern Australia. PhD Thesis, Universityof New England, Armidale, NSW.

Blatt H, Tracey RJ (1996) ‘Petrology: igneous, sedimentary andmetamorphic.’ (W.H. Freeman and Co.: London)

Bormann FH, Likens GE (1979) ‘Pattern and process in a forestedecosystem.’ (Springer: Berlin)

Bratton SP (1976) The response of understorey herbs to soil depthgradients in high and low diversity communities. Bulletin of the TorreyBotanical Club 103, 165–172. doi: 10.2307/2484490

Christensen PE, Kimber PC (1975) Effects of prescribed burning on theflora and fauna of south-west Australian forests. Proceedings of theEcological Society of Australia 9, 85–106.

Clarke PJ (2002) Habitat insularity and fire response traits: evidence from asclerophyll archipelago. Oecologia 132, 582–591.doi: 10.1007/s00442-002-0962-0

Clarke PJ, Knox KJE (2002) Post-fire response of shrubs in the tablelandsof eastern Australia: do existing models explain habitat differences?Australian Journal of Botany 50, 53–62. doi: 10.1071/BT01055

Florence RG (1996) ‘The ecology and silviculture of eucalypt forests.’(CSIRO Publishing: Melbourne)

Gale SJ, Hoare PG (1991) ‘Quaternary sediments: petrographic methods forthe study of unlithified rocks.’ (Belhaven Press: London)

Gary GJ, Morrison DA (1995) Effects of fire frequency on plantspecies composition of sandstone communities in the Sydney region:combinations of fire intervals. Australian Journal of Ecology 20,418–426. doi: 10.1111/j.1442-9993.1995.tb00558.x

George A (2002) ‘The long dry.’ (Four Gables Press: Kardinya, WA)Gill AM (1981) Adaptive responses of Australian vascular plant species

to fires. In ‘Fire and the Australian biota’. (Eds AM Gill, RH Groves,IR Noble) pp. 243–271. (Australian Academy of Science: Canberra)

Gill AM, Bradstock R (1992) A national register for the fire responses ofplant species. Cunninghamia 2, 653–660.

Page 13

748 Australian Journal of Botany A. Benwell

Gill AM, Groves RH (1981) Fire regimes in heathlands and their plant-ecological effects. In ‘Ecosystems of the world. Vol. 9B. Heathlandsand related shrublands: analytical studies’. (Ed. RL Specht) pp. 61–84.(Elsevier: Amsterdam)

Grime JP (1979) ‘Plant strategies and plant processes.’ (Wiley: New York)Grime JP (1993) Stress, competition, resource dynamics and vegetation

processes. In ‘Plant adaptations to environmental stress’. (Eds L Fowden,T Mansfield, J Stoddart) pp. 45–63. (Chapman and Hall: New York)

Grubb PJ (1977) The maintenance of species-richness in plant communities:the importance of the regeneration niche. Biological Review 52,107–145.

Hansen A, Pate JS, Hansen AP (1991) Growth and reproductiveperformance of a seeder and resprouter species of Bossiaeaas a function of plant age after fire. Annals of Botany 67, 497–509.

Hays WL (1981) ‘Statistics.’ 3rd edn. (Rinehart and Winstrom: New York)Hnatiuk RJ, Hopkins AJM (1980) Western Australian species rich kwongan

(sclerophyllous shrubland) affected by drought. Australian Journalof Botany 28, 573–585. doi: 10.1071/BT9800573

Hopper SD (2000) Creation of conservation reserves and managing fire ongranite outcrops. Journal of the Royal Society of Western Australia 83,173–186.

Hopper SD, Brown AP, Marchant NG (1997) Plants of Western Australiangranite outcrops. Journal of the Royal Society of Western Australia 80,141–158.

Hunter JT (1999) Floristics and biogeography of the granitic outcrop flora ofthe New England batholith of Eastern Australia. PhD Thesis, Universityof New England, Armidale, NSW.

Hunter JT (2003) Persistence on inselbergs: the role of obligate seeders andresprouters. Journal of Biogeography 30, 497–510.

Hunter JT, Clarke PJ (1998) The vegetation of granitic outcrop communitiesof the New England batholith of Eastern Australia. Cunninghamia 5,547–618.

Huston MA (1994) ‘Biological diversity, the coexistence of species onchanging landscapes.’ (Cambridge University Press: Cambridge, UK)

Keith D (1996) Fire-driven extinction of plant populations: a synthesis oftheory and review of evidence from Australian vegetation. Proceedingsof the Linnean Society of NSW 116, 3776.

Kershaw AP (1985) An extended late Quaternary vegetation record fromnorth east Queensland and its implications for the seasonal tropicsof Australia. Proceedings of the Ecological Society of Australia 3,179–189.

Main YB (1997) Granite outcrops: a collective ecosystem. Journal of theRoyal Society of Western Australia 80, 113–122.

Moore SM, McCabe GP (1999) ‘Introduction to the practice of statistics.’(WH Freeman & Co.: New York)

Naveh Z (1975) The evolutionary significance of fire in the Mediterraneanregion. Vegetatio 29, 199–208. doi: 10.1007/BF02390011

Pate JS (1993) Structural and functional responses to fire and nutrient stress:case studies from the sandplains of South-West Australia. In ‘Plantadaptation to environmental stress’. (Eds L Fowden, T Mansfield,J Stoddart) pp. 189–205. (Chapman and Hall: London)

Porembski S, Barthlott W (2000) ‘Inselbergs: biotic diversity of isolatedrock outcrops in tropical and temperate regions.’ (Springer: Berlin)

Siddiqui MY, Carolin RC, Myerscough PJ (1976) Studies in the ecology ofcoastal heath in New South Wales. III. Regrowth of vegetation after fire.Proceedings of the Linnean Society of New South Wales 101, 53–63.

Singh G, Geissler EA (1985) Late Cainozoic history of vegetation, fire,lake levels and climate at Lake George, New South Wales, Australia.Philosophical Transactions of the Royal Society of London 311,379–447. doi: 10.1098/rstb.1985.0156

Sousa WP (1984) The role of disturbance in natural communities. AnnualReview of Ecology and Systematics 15, 353–391.doi: 10.1146/annurev.es.15.110184.002033

Tabachnick BG, Fidell LS (1989) ‘Using multivariate statistics.’ (Harper andRow: New York)

Walker J, Hopkins MS (1990) Vegetation. In ‘Australian soil and landsurvey’. (Eds RC McDonald, RF Isbell, JG Speight, J Walker,MS Hopkins) pp. 44–67. (Inkata Press: Melbourne)

Wark MC, White MD, Roberston DJ, Marriott PE (1987) Regenerationof heath and heath woodland in the north-eastern Otway Rangesfollowing the wildfire of February 1983. Proceedings of the Royal Societyof Victoria 99, 51–88.

White PS (1979) Pattern, process and natural disturbance in vegetation.Botanical Review 45, 229–299.

Williams JE, Gill AM (1995) ‘The impact of fire regimes on native forestsin Eastern New South Wales. Forest issues 1.’ (NSW National Parks andWildlife Service: Hurstville)

Wooller SJ, Wooler RD, Brown KL (2002) Regeneration by three species ofBanksia on the south coast of Western Australia in relation to fire interval.Australian Journal of Botany 50, 311–317. doi: 10.1071/BT01078

Wyatt R, Fowler N (1977) The vascular flora and vegetation of the NorthCarolina granite outcrops. Bulletin of the Torrey Botanical Club 104,245–253. doi: 10.2307/2484306

Yates CJ, Hobbs RJ, Bell RW (1994) Landscape scale disturbances andregeneration of semiarid woodlands of southwestern Australia. PacificConservation Biology 1, 214–221.

Yates CJ, Hopper SD, Brown A, van Leeuwen S (2003) Impact of twowildfires on endemic granite outcrop vegetation in Western Australia.Journal of Vegetation Science 14, 185–194.doi: 10.1658/1100-9233(2003)014[0185:IOTWOE]2.0.CO;2

Manuscript received 16 March 2007, accepted 28 May 2005

http://www.publish.csiro.au/journals/ajb