How hot? How often? Getting the fire frequency and timing right for optimal management of woody...

How hot How often Getting the fire frequency and timingright for optimal management of woody cover and pasturecomposition in northern Australian grazed tropical savannasKidman Springs Fire Experiment 1993ndash2013

Robyn A CowleyAD Mark H HearndenA Karen E JoyceB Miguel Tovar-ValenciaBTrisha M CowleyA Caroline L PettitA and Rodd M DyerC

ANorthern Territory Department of Primary Industry and Fisheries PO Box 1346 Katherine NT 0851 AustraliaBCharles Darwin University Darwin NT 0909 AustraliaCACIAR GPO Box 1571 Canberra ACT 2601 AustraliaDCorresponding author Email robyncowleyntgovau

Abstract A long-term (1993ndash2013) experiment in grazed semiarid tropical savannas in northern Australia tested theimpact of varying the frequency (every 2 4 and6years) and season (JunendashEDSversusOctoberndashLDS)offire comparedwithunburnt controls onwoody cover and pasture composition in grassland and openwoodlandOver an 18-year period woodycover increased by 4 (absolute) in the woodland even with the most severe (ie frequent late dry season) fire treatmentsWith less severe or nofirewoodycover increased by12ndash17 In the grasslandwoodycover remained staticwhen subjectedto LDS fires every 2 or 4 years but increased by 3ndash6 under other fire treatments and by 8when unburnt Major shifts inunderstorey species composition occurred at both sites regardless of fire regime The effect of fire on herbage mass andcomposition was compounded by higher grazing after fires The herbage mass of perennial grasses declined and that ofannual grasses and forbs increased following early or frequent fires Brachyachne convergens Gomphrena canescens andFlemingiapauciflora increased in response tofirewhileAristida latifolia andHeteropogoncontortusdecreased Four-yearlyLDS fire provided the most effective management of woody cover and pasture composition Although EDS fire isrecommended for biodiversity management and reducing greenhouse gas emissions in wet tropical savannas on grazedpastoral land it can promotewoodland thickening and pasture degradationOptimalfiremanagement therefore depends onvegetation type land use and the prevailing seasonal timing and frequency of fire

Additional keywords fire management fire season rangeland management time since fire

Received 5 March 2014 accepted 13 August 2014 published online 24 September 2014

Introduction

Potential woody plant cover within savanna ecosystems isbound by the abiotic drivers of rainfall nutrients and CO2

(Williams et al 1996 Bond and Midgley 2000 Sankaran et al2005Mills et al 2006 Bond 2008 February et al 2013)Withinthis envelope of environmental potential savanna structure isalso influenced by modifiers such as woody plant consumersfire and herbivores (Archer 1995 Staver et al 2009 Sankaranet al 2013) Across uncleared Australian savannas the relativeimportance of controlling factors varies regionally with climate(Fensham and Fairfax 2003 Fensham et al 2005) grazing bydomestic livestock and changed fire management (Burrows et al2002 Sharp and Whittaker 2003 Crowley et al 2009) beingimportant in different areas over the last century

Removal of pasture by grazing promotes an increase inwoodycover by reducing competition between the under- and overstorey

layers (February et al 2013) and by reducing incidence andintensity of fire through reduction in the amount of fuel Even inthe absence of grazing woody cover naturally fluctuates throughtime generally increasing between major mortality eventsdriven by disturbance or drought (Fensham et al 2005) There isincreasing evidence that rising concentrations of atmosphericCO2 may be shifting the balance towards increasing woodinessacross savannas globally (Kgope et al 2010) Counterbalancingthis if climate change leads to increasing severity and frequencyof drought vegetation thinning could result (Fensham et al2009)

Increases in woody vegetation are most often associated withgrazing by livestock and the exclusion of fire particularly onfertile (Archer 1995 Burrows et al 2002) or seasonally floodedareas (Crowley and Garnett 1998 Sharp and Whittaker 2003)In ungrazed savannas with high frequencies of fire woody cover

Journal compilation Australian Rangeland Society 2014 wwwpublishcsiroaujournalstrj

CSIRO PUBLISHING

The Rangeland Journal 2014 36 323ndash345httpdxdoiorg101071RJ14030

has more typically been stable or declined (Russell-Smith et al2010 2012 Murphy et al 2014) Contrasting trajectories ofwoody cover with different fire and grazing management havebeen evident in the Victoria River District of the NorthernTerritory Here some grazed landscapes have undergonewoodland thickening since settlement (Sharp and Whittaker2003 Lewis et al 2010) while in other areas woody cover hasbeen relatively stable or even declined (Fensham and Fairfax2003 Lewis et al 2010) Although both grazing and reducedfrequency of fire on productive pastoral land types have beenimplicated in woodland thickening in the Victoria River District(Lewis 2002 Sharp andWhittaker 2003) potential woody coverin the regionmay also be rising in response to the changing abioticdrivers of increasing rainfall since the 1970s (Shi et al 2008) andrising atmospheric CO2 concentrations

Woodland thickening reduces pasture growth (McIvor andGardener 1995 Jackson and Ash 1998 Scanlan 2002) andsubsequent carrying capacity for livestock production (Dyerand Stafford Smith 2003) although the nutritive value ofpasture is often enhanced under trees (Jackson and Ash 2001Scanlan 2002)

Fire is the only economically viable tool available to managewoody cover (Rolfe 2002 Landsberg et al 2011) and ispreferable to clearing which has adverse impacts on biodiversity(McAlpine et al 2002) ecosystem function (Ludwig andTongway 2002) and greenhouse gas emissions (Burrows et al2002) Fire is also used extensively for other purposes on pastoralproperties as reviewed elsewhere (Tothill 1971 Pressland 1982Dyer et al 2001 2003)

Reducing grazing pressure to allow pasture species tooutcompetewoodyplants or to limit their rate of increase is oftenineffective (Archer 1995) While browsing of woody plants canrestrict their growth (Scholes and Archer 1997 Sankaran et al2013 Staver and Bond 2014) plants prone to woody thickeningtend to be unpreferred (Crowley et al 2009) and in Africansavannas the combination of browsing and fire is required tosuppress tree density (Staver et al 2009)

Fire decreases rates of woody growth (Murphy et al 2010)and stem survival (Williams et al 1999 Murphy et al 2014)However the adaptive traits to fire such as resprouting confervery lowmortality onwoodyplants evenunder regimes of intensefire (Bond and Midgley 2001 Dyer 2001) Hence fire tends toreduce the size and cover ofwoody plants temporarily rather thanaffect their density (Higgins et al 2007)

Fire can also influence composition of the understoreyFire consumes aboveground biomass of herbaceous speciesindiscriminately (Bond and Keeley 2005) In the absence ofgrazing this can provide a competitive advantage to grazing-sensitive species because their less-preferred neighbours are alsoconsumed by fire Burnt pasture however attracts herbivores(Andrew 1986a Southwell and Jarman 1987 Vermeire et al2004) and if burnt patches are relatively small they canbe subjectto very high grazing after fire Hence fire can indirectly promotespecies that avoid grazing through shorter lifespan (temporalavoiders ndash annuals) prostrate or short habit (spatial avoiders) orlow palatability (biochemical avoiders) (Briske 1998 Landsberget al 1999) at the expense of perennial tussock grass speciesthat are more susceptible to grazing Fire can also act to increasegrazing on less-preferred species afterfire (Dyer et al 2001)when

cattle are less selective about the species that they graze (Andrew1986b)

In the light of recommendations for a shift from late to earlydryseason fire to abate carbon emissions and enhance biodiversityconservation (Russell-Smith et al 2013 Walsh et al 2014) thisstudy seeks to improve understanding of the impacts of fireregimes in grazed savannas with a focus on impacts on woodycover and pasture condition Its 20-year duration allowsassessment of the interactions with rainfall and grazing

Methods

Study site

This studywas established in 1993on theVictoriaRiverResearchStation also known as lsquoKidman Springsrsquo 400 km south ofDarwin Northern Territory Australia in the semiarid tropicalsavannas The site has a July to June median rainfall of 742mm(wwwbomgovau rainfall station No 14847 accessed 23August 2014)Consistentwith the increase in the rainfall of north-western Australia since the 1970s (Shi et al 2008) the medianrainfall during the study of 770mm exceeded that of the previous21 years (640mm) Between-year variability in rainfall wasgreater in the studyrsquos last 10 years than during the first 10 years

Twoexperimental siteswere established4 kmapart in separategrazed (by Bos indicus cattle) paddocks of native pasturesin woodland and grassland (Fig 1 Table 1) At each site 16experimental plots (~26 ha in area) were arranged in a 4 4 gridwith each plot separated by cleared firebreaks Two plots were

(a)

(b)

Fig 1 Aerial photograph of the (a) woodland and (b) grassland sites inJune 2013

324 The Rangeland Journal R A Cowley et al

randomly allocated to each of six fire treatments [E2 E4 E6L2 L4 L6 where E and L signify early dry season (EDS) andlate dry season (LDS) fire and 2 4 and 6 signify intervals betweenfires of 2 4 and 6 years] and four plots were assigned as unburntcontrols The fires of EDS were lit in June and those of LDS inOctober

The average frequency of fires on the plots since 1993 wasclose to the target intervals of 2 (burnt in 47 of years) 4 (26)and 6 (16) years (Table 2) although poor seasonal conditionssometimes prevented burning In addition to experimental fireplotsfireswere usually lit in surrounding parts of the paddocks toreduce post-fire grazing on the unfenced experimental fire plotsbut this was not always carried out

Data collection

Stocking rates were calculated on a wet season basis (OctoberndashSeptember) yearly for each experimental paddock based on thenumber of cattle recorded in the yards in the October and Maymusters

Fuel load (dry matter (DM) of herbage mass measured byBotanal see below) was assessed across the sites in the daysbefore burning Dry bulb temperature and relative humidity weremeasured on site with field instruments before each fire Flame

height (averageminimumandmaximum)was visually estimatedduring the burn

Total herbage mass of DM herbage mass of DM of speciesground cover and grazing intensity were visually estimated usingBotanal (Tothill et al 1978) in June and October annually from1994 to 2001 then in June biennially from 2003 to 2013 Onlyplots due to be burned were assessed in October 2003 and 2005and all plots were assessed again in October 2013 Visualestimates were calibrated against estimates of herbage mass ofDM from 10 quadrats cut on each sampling day Grazing scorethe proportion of quadrat herbage mass removed by grazing[0- (0) 1- (1ndash5) 2- (6ndash25) 3- (26ndash50) 4- (51ndash75)5- (76ndash100)] was visually assessed from 2007 onwardsBotanalmeasurementswere taken in 15 1-m2 quadrats positionedat ~8-m intervals along four permanent parallel northndashsouthtransects in each plot

Canopy and basal area cover of woody plants were assessedusing aBitterlich gauge at every pasture quadrat in 2009 Obliqueaerial images were used to assess woody canopy cover in 1995and 2013 Individual plot photos were geo-referenced and abinary classification routine was created for these images toextract the percentage cover of woody vegetation (K Joyce perscomm) Change inwoody cover is presented as the total absolutechange in per cent cover between 1995 and 2013 and the average

Table 1 Characteristics of the experimental siteSoil description based on Cobiac (2001)

Woodland Grassland

Paddock name Conkerberry Rosewood WestRelevant sites from Cobiac (2001) 11 9 and 10Land system (Stewart et al 1970) Humbert ArgyleDominant overstorey species Corymbia terminalis Eucalyptus pruinosa Carissa

lanceolata Hakea arborescensTerminalia volucris Bauhinia cunninghamii

Soil Calcarosol (calcareous red earth) Vertosol (alluvial grey cracking clay)Soil depth 04m gt2mSoil texture (0ndash20 cm) 36 clay 11 silt 53 sand 50 CaCo3 nodules

between 02ndash04m49 clay 11 silt 40 sand

Soil water-holding capacity (to 1m) 73mm 247mm

Table 2 History of fire on the experiment from 1993 to 2013Six fire treatments [E2 E4 E6 L2 L4 L6 where E and L signify early dry season (EDS) and late dry season (LDS) fire and 2 4 and 6 signify intervals between

fires of 2 4 and 6 years] were applied to woodland and grassland sites Shading indicates the season and year that plots were burnt

Treatment 1993 1994 1995 1996 1997 1999 2001 2003 2005 2007 2009 2011 2013 Number of burns years burnt asE L E L E L E L E L E L E L E L E L E L E L E L E L since 1993 of June 2013

Control 1A 2A 01 5E2 B 10 47E4 C 5 26E6 B 4 16L2 B 10 47L4 C 5 26L6 B 4 16

A1 2 accidental fires in one of the four control plots due to fires from treatment plots spotting into neighbouring control plot in late burns (12001 Woodland22009 Grassland site)

B2007 fires unable to be implemented due to low fuel loads and high moisture content following out-of-season rainfall in JuneCFires not implemented due to human error

Fire impacts in a grazed tropical savanna The Rangeland Journal 325

annual change in per cent cover (total cover change18 years) tofacilitate comparison with other studies

Statistical analysis

For each herbagemass cover and grazing scoremeasure a singlevalue was derived for each plot by averaging values from all60 quadrats This provided two values per treatment per siteeach year and four values for controls Where requireddependent variables were transformed to meet the assumptionsof ANOVA The impact of fire was analysed using twocomplementary approaches First plots were grouped andanalysed according to the time since the last fire (TSF) regardlessof season or frequency of fire Second plots were analysed inrelation to the treatments of season and frequency of fire

Time since fire

A single factor ANOVA was used to assess the impact ofTSF on herbage mass and ground cover in a subset of years withat least three or more times since fire (1996 1997 2001 20072011 and 2013) Tukeyrsquos HSD was used for post-hoc pair-wisecomparisons of means and HSD tests for unequal N were usedwhere sample size varied between groups Non-parametricKruskalndashWallis ANOVA on ranks was used to assess the impactof TSF on grazing intensity

Season and frequency of fire

A factorial ANOVAwith a hanging control was used to assessthe effect of fire treatment on woody cover and grazing ineach year Dunnettrsquos test was used to compare main effects withcontrols Post-hoc pair-wise comparisons of means were carriedout using Fisherrsquos protected lsd tests

For each year when adequate data were available impact ofseason and frequency of fire and year on herbage mass groundcover and herbage mass of major species and functional groupswere analysed using a within-effects model (repeated-measures)ANOVAwith a balanced control for all years from 1994 to 2013Because variables often changed through time in response tofire treatments and because there was often a relatively high levelof variability between the two replicate plots direct effectswere rarely significant Rather a significant year treatmentinteraction may indicate an effect due to the fire treatments Forthis reason factorial ANOVA was used to assess frequency andseason effects of fire for each year separately A single factorANOVAwas also used to test for the effect of any fire (fire versusno fire) for each year separately Bonferonni tests were used forpost-hoc pair-wise comparisons of means Because 6-yearlytreatments were rarely significantly different from controls theyare excluded from the figures to improve readability

Results

Site vegetation

Total herbage mass and ground cover in the control plotsfluctuated annually and were initially lower in the woodland site(Fig 2a b) but converged through time Low total rainfall in2003 coupled with a late start to the wet season and a late start tothe wet season in 2007 with 115mm lsquoout of seasonrsquo rainfall inMayndashJune led to an exceptionally low herbage mass and groundcover in those years

In the control plots of the woodland site ground-layerdominance changed from annual and short-lived perennialgrasses to perennial grasses with an increase in Heteropogoncontortus from 8 to 73 of the herbage mass occurringmostlyafter 1999 (Fig 2c) Species preferred by livestock decreasedfrom 73 to 28 of the herbage mass in the control plots of thegrassland site between June1994and June2013Aristida latifoliaincreased at the expense of Chrysopogon fallax and Iseilemaspp (Fig 2d)

Stocking rates

Stocking rate averaged 105 adult equivalents (AE) kmndash2 in thewoodland site and 125AE kmndash2 in the grassland site These ratesapproximated to or were lower than the estimated long-term safecarrying capacities (LTCC Cowley and Bryce 2004 woodland11AEkmndash2 grassland 15AE kmndash2) Stocking rates variedannually but rarely exceeded LTCC (Fig 3) The exceptionswere in the grassland site in 2002 in both sites in the very dry yearof 2003 and in the woodland site in 2007

Fire characteristics

Weather conditions measured at the site immediately beforeburning were generally cooler and more humid preceding fires inthe EDS treatments than in the LDS treatments The averageair temperature before fires in the EDS treatments was 308C(range 21ndash43) versus 388C (range 33448C) for fires in the LDStreatments The average relative humidity before fires was 29(12ndash43) forfires in theEDS treatments versus 22 (9ndash45) forfires in the LDS treatments Averaged across all dates sites andtreatments flame height of fires in the LDS treatments was 1mhigher than those in the EDS treatments (3m versus 2m) andflame height on 4- and 6-yearly fires was 1m higher than for2-yearly fires (3m versus 2m)

Fuel load of herbage in the woodland site was generally lowerthan that in the grassland site (woodland site treatment E21557 kgDMhandash1 treatment E4 1946 kgDMhandash1 treatment E62010 kgDMhandash1 grassland site treatment E2 1895 kgDMhandash1treatment E4 2249 kgDMhandash1 treatment E6 2732 kgDMhandash1)In some years (1993 in the woodland site 2003 and 2007 inboth sites) very low fuel loads (lt1000 kg DM handash1) preventedeffective fires Fuel loads on the EDS treatments were generallya few hundred kg DM handash1 greater than the fuel loads on the LDStreatments

Time since fire

Woody cover

In 1995 woody cover increased with TSF in the grasslandsite but not in the woodland site (Table 3) In 2009 there was atendency for lower woody cover on more recently burnt plotsand by 2013 plots which had not been burnt for a long time hadapproximately double the woody cover of more recently burntplots

Grazing intensity herbage mass and ground cover

Grazing intensity was not affected by TSF at the woodlandsiteOn thegrassland site in 2007 and2013whenherbage growthwas low plots burnt in the previous 2 years had higher grazingintensity compared with plots burnt less recently (Table 4) Total


0

1000

2000

3000

4000

0

20

40

60

80

100

0

20

40

60

80

0

20

40

60

Woodland

Grassland

Woodland

Grassland

Brachyachne convergens

Enneapogon polyphyllus

Dichanthium fecundumHeteropogon contortus

Aristida latifolia

Chrysopogon fallaxDichanthium fecundumIseilema spp

Her

bage

mas

s (k

g D

M h

andash1)

Gro

und

cove

r (

)

of h

erba

ge m

ass

o

f her

bage

mas

s

(a)

(b)

(c)

(d)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

Fig 2 (a)Mean total herbagemass (b) ground cover (c) herbage composition inwoodland site and (d) herbage composition ingrassland site of control plots in June of each year (Bars represent se of mean)


herbage mass and ground cover were sometimes lower in thefirst year after fire compared with plots subject to fire 2 ormore years ago The effect of fire on herbage mass was morepronounced in the grassland site with herbage mass still lower(and grazing intensity higher) 2 years post-fire in 2013

Herbage mass of functional groups and species

Changes in the herbage mass of functional groups and specieswere observed up to 2 years post-fire The per cent herbage massof dicots (and occasionally absolute herbage mass of dicots) wasoften about twice as high in the first or second year following firecompared with plots burnt less recently However trends wereonly marginally statistically significant (Table 4) In 1996 and1997 for the woodland site and in 1997 for the grassland siteherbagemass of annual grasseswas lower in thefirst year afterfirecompared with plots burnt less recently However in 2001 2011and 2013 this pattern had reversed so that herbagemass of annualgrasses and per cent herbage mass of annual grasses was oftenhigher 2 years after fire Plots burnt in the previous year or twooften had lower herbagemasses of perennial grasses (and per centherbage mass of perennial grasses) compared with those burntless recently Initially this pattern appliedonly to thegrassland site

but by 2011 per cent herbage mass of perennial grasses waslowest on the most recently burnt plots on the woodland site

Herbage mass and per cent herbage mass of Brachyachneconvergens was often highest on the most recently burnt plots(Table 4) Herbage mass of Enneapogon polyphyllus wasinitially lowest on plots burnt in the previous year in 1996 but by2001 the pattern was reversed with the lowest herbage massof E polyphyllus on plots burnt more than 18 years ago althoughdifferences were not statistically significant in later years Theherbage mass of Iseilema spp did not respond consistently toTSF The herbage mass of Dichanthium fecundum Sehimanervosum C fallax and A latifolia were occasionally lower onmore recently burnt plots but the large variability in herbagemassbetween replicate plots meant that the effect of TSF was rarelystatistically different Per cent herbage mass of Heteropogoncontortuswas initially lowest on plots burnt 2 years previously in2001 but differenceswere not significant until 2011 The herbagemass of Gomphrena canescens was highest on plots in thewoodland site burnt 2 years previously in 2011 and 2013 Percentherbage mass of Flemingia pauciflorawas highest on plots in thegrassland site burnt 2 years previously in 2013


Grazing

Grazing intensity in the woodland site was not affected by firetreatment In the grassland site 2- and 4-yearly burnt plots hadhigher grazing intensities in 2007 and 2013 compared withcontrol plots (Table 5) reflecting the shorter time since fire ofthese treatments In 2011 the E2 treatment was more heavilygrazed than the L2 treatment Grazing was not affected by fire inthe grassland site in 2009 (not shown)

Woody cover

In 1995meanwoody cover differed between sites (woodland56 grassland 36 ANOVA P lt 001) Despite fire regimesbeing established between 12 and 18 months before 1995there was little difference between control and burnt plots(Fig 4a b) although in the woodland site EDS treatments hadsignificantly higher cover thanLDS treatments (Tukeyrsquospost-hoctest P lt 005)

Woody cover increased between 1995 and 2013 for alltreatments on the woodland site and most treatments on the

0

5

10

15

20

25

Sto

ckin

g ra

te (

AE

km

ndash2)

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013

Fig 3 Actual stocking rate in thewoodland (grey dotted line) and grassland(black solid line) sites The safe long-term carrying capacity for the woodlandsite (solid grey bar) and the grassland site (solid black bar) are also given[1 animal equivalent (AE) is a 450 kg steer]

Table 3 ANOVA and mean canopy cover across treatments ( se of mean) of the effect of time since fireValues followed by different letters denote Fishersrsquos lsd significant differences within a site W Woodland site G grassland site

Year and variable Site F-value P Time since fire

1995 (aerial assessment) ndash ndash ndash 1 year (n= 10) 2 years (n= 2) gt2 years (n= 4)Canopy cover () Woodland 034 072 57 plusmn 079 43 plusmn 128 57 plusmn 055

Grassland 434 003 23 plusmn 033a 54 plusmn 171ab 60 plusmn 208b2009 (ground assessment) ndash ndash ndash 4 years (n= 8) 8 years (n= 5W4G) 16 years (n= 3W4G)Tree basal area (m2 handash1) Woodland 292 009 26 plusmn 029a 36 plusmn 036b 34 plusmn 047ab

Grassland 359 006 09 plusmn 025a 09 plusmn 035a 20 plusmn 035bCanopy cover () Woodland 402 004 160 plusmn 126a 204 plusmn 159b 217 plusmn 205b

Grassland 172 021 77 plusmn 163 91 plusmn 231 129 plusmn 2312013 (aerial assessment) ndash ndash ndash 2 years (n= 8) 12 years (n= 4W5G) 20 years (n= 3)Canopy cover () Woodland 1316 00007 117 plusmn 137a 222 plusmn 156b 191 plusmn 220b

Grassland 352 006 53 plusmn 139a 75 plusmn 234a 143 plusmn 460b


Table 4 ANOVA and means of ground cover total herbage mass (kg DM handash1) and herbage mass and of herbage mass of functional groups andspecies showing the effect of time since fire

Means (se ofmean) across all plots are shown for each time sincefireMeans followedby different letters are significantly differentUnequalHSDwhere unequalNunlessonlymarginally significant otherwiseTukeyrsquosHSD^KruskalndashWallisANOVAandmedianWWoodland siteG grasslandsiteOnly instanceswhere

there was a trend with TSF are shown


1996 ndash ndash ndash 1 year (n= 4) 2 years (n= 6) 3 years (n= 2) 13 years (n= 4)Total herbage mass W 962 lt001 1079 (339)a 1367 (973)a 1442 (1768)ab 1899 (1354)b

G 478 lt005 1628 (578)a 2389 (1946)ab 2978 (467)b 2463 (3195)abGround cover () W 826 lt001 40 (17)a 50 (25)ab 51 (31)ab 61 (40)b

G 153 lt0001 54 (17)a 73 (21)b 79 (64)b 72 (26)b dicots W 115 036 23 (102) 17 (28) 6 (28) 12 (40)Annual grasses W 406 lt005 423 (313) 757 (828) 941 (558) 851 (1633)

G 126 033 222 (166) 190 (354) 126 (388) 145 (365)Perennial grasses G 459 lt005 1263 (835)a 2033 (2022)ab 2719 (321)b 2199 (3662)ab perennial grasses G 480 lt005 77 (22)a 84 (20)ab 91 (02)b 87 (33)abBrachyachne convergens W 207 016 204 (467) 148 (574) 240 (203) 71 (230) B convergens W 342 005 20 (50)a 14 (65)ab 22 (10)a 4 (11)bDichanthium fecundum W 014 093 27 (99) 50 (205) 54 (541) 390 (3797) D fecundum W 017 091 3 (11) 4 (13) 4 (37) 17 (158)D fecundum G 186 019 133 (492) 413 (910) 738 (4031) 462 (2537)Enneapogon polyphyllus W 372 lt005 207 (317)a 578 (1032)ab 679 (907)ab 754 (1777)b E polyphyllus W 224 014 19 (24) 41 (57) 45 (108) 42 (112)Gomphrena canescens W 073 055 97 (701) 105 (372) 38 (377) 27 (140) G canescens W 080 052 9 (65) 7 (27) 3 (28) 1 (07)

1997 ndash ndash ndash 1 year (n= 2) 2 years (n= 4) 3 years (n= 6) 14 years (n= 4)Total herbage mass W 085 049 1644 (499) 1853 (1302) 1783 (819) 1918 (991)

G 735 lt001 1735 (2597)a 3035 (2194)ab 3608 (1892)b 3140 (3010)abGround cover () W 066 059 67 (36) 70 (40) 68 (14) 65 (19)

G 519 lt005 63 (43)a 75 (16)ab 78 (13)b 73 (38)ab dicots G 159 024 9 (04) 4 (10) 4 (12) 4 (14)Annual grasses G 234 013 388 (651) 920 (679) 779 (1322) 525 (1706)Perennial grasses W 101 042 410 (2086) 578 (1596) 451 (1068) 777 (1989)

G 285 008 1198 (1880) 2004 (1844) 2645 (2932) 2481 (4072)B convergens W 229 013 727 (1315) 558 (1666) 347 (1560) 198 (849) B convergens W 234 012 44 (51) 30 (86) 22 (104) 10 (42)D fecundum G 102 042 123 (917) 288 (1139) 518 (1420) 402 (1731)D sericeum G 114 037 119 (450) 89 (468) 65 (229) 29 (138) D sericeum G 353 lt005 8 (23)a 3 (15)ab 2 (07)b 1 (04)bE polyphyllus W 071 056 274 (506) 296 (1032) 471 (655) 531 (1542)G canescens W 110 039 45 (458) 273 (1232) 263 (883) 107 (619) G canescens W 115 037 3 (27) 16 (69) 14 (47) 6 (36)Iseilema spp G 330 006 205 (1235)a 785 (1097)b 701 (1458)b 449 (1998)abSehima nervosum G 070 057 17 (166) 344 (2351) 265 (1125) 436 (1574) S nervosum G 055 066 1 (11) 10 (68) 7 (28) 12 (53)

2001 ndash ndash ndash 2 years (n= 4) 4 years (n= 4) 6 years (n= 4) 18 years (n= 4)Total herbage mass W 216 015 2630 (1100) 2774 (691) 2222 (2649) 2516 (1213)

G 060 063 2641 (1060) 2846 (2619) 2803 (1779) 2553 (1222)Ground cover () W 216 015 79 (22) 78 (17) 68 (62) 7 (18)

G 042 074 82 (32) 77 (54) 78 (53) 75 (33)Annual grasses W 335 006 1172 (1547) 1039 (1701) 692 (2065) 541 (975)

G 478 lt005 532 (600)a 205 (333)ab 200 (532)b 243 (1009)ab annual grasses W 240 012 48 (52) 40 (55) 35 (108) 23 (44)

G 412 lt005 21 (26)a 7 (14)ab 7 (20)b 10 (48)abD fecundum W 054 066 136 (648) 242 (1530) 305 (705) 394 (3392) D fecundum W 061 062 5 (23) 9 (53) 12 (23) 14 (121)D fecundum G 185 019 187 (602) 506 (1063) 426 (1100) 478 (1777) D fecundum G 172 022 7 (22) 19 (54) 15 (39) 19 (77)E polyphyllus W 165 023 623 (719) 733 (1897) 448 (1983) 301 (1220) E polyphyllus W 113 038 25 (27) 27 (68) 22 (98) 13 (54)G canescens W 177 021 40 (81) 35 (170) 18 (108) 8 (49) G canescens W 119 035 2 (03) 1 (08) 1 (06) 0 (02)

(continued next page)


Table 4 (continued )


Heteropogon contortus W 103 041 38 (31) 45 (35) 40 (72) 54 (104)Iseilema spp G 409 lt005 439 (668)a 126 (280)ab 122 (432)b 182 (1032)ab Iseilema spp G 415 lt005 17 (28)a 4 (08)ab 4 (15)b 8 (49)ab

2007 ndash ndash ndash 2 years (n= 8) 6 years (n= 5W4G) 24 years (n= 3W4G) ndash

Grazing score^ W 34 018 23 15 15 ndash

G 901 lt001 38a 32ab 26b ndash

Total herbage mass W 140 028 852 (534) 1100 (1776) 939 (1212) ndash

G 293 009 572 (1153) 994 (2176) 948 (961) ndash

Ground cover () W 204 017 32 (22) 40 (49) 39 (58) ndash

G 423 lt005 23 (25)a 31 (50)ab 35 (16)b ndash

Dicots W 393 0046 91 (171) 38 (103) 44 (143) ndash

dicots W 591 lt001 13 (21)a 4 (14)b 5 (20)ab ndash

G 297 008 7 (20) 4 (08) 2 (11) ndash

Perennial grasses G 369 lt005 537 (1156) 965 (2179) 925 (1023) ndash

perennial grasses G 308 008 91 (19) 96 (09) 97 (13) ndash

Aristida latifolia G 311 008 59 (124) 176 (538) 308 (1585) ndash

A latifolia G 083 046 14 (38) 24 (98) 36 (185) ndash

D fecundum G 088 044 54 (209) 160 (838) 127 (635) ndash

2011 ndash ndash ndash 2 years (n= 4W5G) 6 years (n= 4) 10 years (n= 5W4G) 28 years (n= 3)Grazing score^ W 299 039 01 01 01 02

G 162 065 13 14 11 13Total herbage mass W 028 083 2885 (1958) 2980 (2839) 3160 (2227) 2919 (2623)

G 245 011 2500 (1262) 2707 (1410) 2952 (1193) 2895 (1613)Ground cover () W 246 011 96 (04) 94 (13) 93 (07) 93 (17)

G 035 079 89 (28) 89 (46) 93 (28) 91 (35)Dicots W 657 lt001 962 (673)a 526 (508)b 536 (990)b 416 (961)b dicots W 413 lt005 34 (41)a 18 (16)ab 18 (41)ab 15 (48)bAnnual grasses W 252 011 488 (604) 332 (727) 312 (468) 259 (406) annual grasses W 180 020 18 (32) 12 (32) 11 (23) 9 (08)Perennial grasses W 165 023 1436 (2876) 2123 (3289) 2314 (3069) 2244 (3272)

G 297 007 1906 (986) 2003 (1798) 2377 (1575) 2362 (1213) perennial grasses W 572 lt001 48 (61)a 70 (41)ab 72 (53)b 76 (47)bA latifolia G 094 045 639 (1416) 499 (1024) 813 (2015) 1004 (3125) A latifolia G 098 043 25 (45) 18 (35) 28 (71) 34 (94)B convergens W 631 lt001 272 (419)a 102 (293)b 104 (317)b 102 (229)ab B convergens W 376 lt005 10 (19)a 3 (09)b 4 (11)b 4 (04)bE polyphyllus W 179 020 178 (102) 153 (548) 144 (307) 68 (261) E polyphyllus W 120 035 7 (10) 6 (24) 5 (14) 3 (10)G canescens W 572 lt001 392 (1197)a 57 (277)b 62 (438)b 60 (452)b G canescens W 442 lt005 14 (47)a 2 (12)ab 2 (17)b 2 (20)abH contortus W 274 009 1219 (3015)a 1992 (2881)b 2028 (2519)b 2060 (2295)b H contortus W 662 lt001 41 (71)a 66 (27)b 63 (43)b 70 (38)b



G 1076 lt001 27a 15b 12b ndash


G 581 lt005 1089 (1079)a 2235 (5178)b 2260 (4620)ab ndash


G 170 022 61 (17) 68 (36) 60 (74) ndash

dicots G 607 lt005 16 (19)a 8 (16)ab 6 (34)b ndash

Annual grasses W 236 013 139 (321) 95 (238) 47 (34) ndash

annual grasses W 270 010 16 (44) 9 (27) 4 (09) ndash

G 549 lt005 14 (24)a 6 (14)b 6 (13)ab ndash

Perennial grasses W 189 019 782 (1205) 1106 (2271) 1267 (2416) ndash

G 654 lt001 762 (1005)a 1937 (5147)b 2016 (4611)b ndash

perennial grasses W 364 006 65 (46) 78 (41) 82 (45) ndash

G 835 lt001 70 (29)a 86 (26)b 88 (39)b ndash



B convergens W 350 006 41 (56)a 30 (42)ab 17 (80)b ndash



grassland site with greater increases in the woodland site(Fig 4c d) The unburned controls in the woodland site hadalmost double the rate of change in woody cover of thegrassland site controls (woodland 081 yearndash1 versus grassland044 yearndash1) Increase in woody cover was lowest with morefrequent LDS treatments on both sites In the woodland site the2-yearly fires and fires on the L4 treatment produced the smallestincrease (38ndash41 and 021ndash023 yearndash1 respectively) whichwas statistically less than for the control plots (Plt 005) whilefires in theE4 treatment andfires on all the 6-yearlyfire treatmentshad similar increases in woody cover (119ndash168 and066ndash093 yearndash1 respectively) to the unburnt plots whichincreased by 145 (081 yearndash1) On average LDS treatments(78 and 044 yearndash1) and 2- and 4-yearly treatments (38and 021 yearndash1 and 82 and 046 yearndash1) had significantlylower increases in cover compared with control plots (P lt 001P lt 0001 and P lt 005 respectively)

In the grassland site L2 and L4 treatments had little changein woody cover in 18 years (ndash011 to 064 and ndash0006 to004 yearndash1) They were the only treatments with a statisticallylower change in cover than the controls which increased by79 (044 yearndash1) (P lt 005) In contrast other fire treatmentsincreased by 31 (treatment E2) to 59 (treatment E6)(017ndash033 yearndash1) On average 2-yearly and LDS treatmentplots had much lower increases in cover than controls (15008 yearndash1 and 16 009 yearndash1 P = 0051 and P lt 005)

Herbage mass and cover

Although herbage mass and cover varied greatlybetween years in response to varying rainfall and stocking ratesthe effects of fire frequency and season were rarely significant(Appendix 1 Table 6 Fig 5a b) Herbage mass tended to beslightly lower at both sites withmore frequent fires andwith earlycompared with late fires or any fires but differences were smalland only statistically different in some years Averaged acrossall years herbage mass on EDS and LDS burnt plots was 1604and 1742 kg handash1 in the woodland and 2104 and 2337 kg handash1 inthe grassland respectively Woodland burnt plots averaged1607 kg handash1 compared with 1870 kg handash1 on the unburntcontrols while on the grassland 2-yearly burnt plots averaged2016 kg handash1 compared with 2361 kg handash1 for 6-yearly burnt andunburnt plots

Herbage mass of functional groups

While analysis of the entire 18 years of data identified year ashaving a significant influence on the herbage mass of functionalgroups the impact of fire frequency was rarely significant andfire season never so (Appendix 1) However significant yearseason and year frequency interactions were identified HencetheANOVAanalyses of functional group yieldwithin yearsweremore likely to be significant in later years This was the case fordicots with a higher proportion of dicots in 2-yearly burnt plots(19) than in controls (10) in the woodland with differencesonly reaching statistical significance in 2011 In the grassland therelative contribution of dicots to herbage mass increased throughtime with greater rises occurring in more frequently burnt and onearly versus late burnt plots The relative contribution of legumesvaried significantlywith yearwith greater rises in burnt plots thanin the controls (Fig 6a)

Herbage mass of annual grasses and per cent of herbagemass crashed in 2003 across all woodland and grassland plots(Fig 6b c) In the woodland herbage mass and percent herbagemass of annual grasses tended to be higher on burnt plotsparticularly 2-yearly burnt plots in the later part of the study Inthe grassland 2-yearly burnt plots had a higher herbage mass ofannual grasses on average than control plots

Herbage mass of perennial grasses was 25 to 40 lower(average all years) on 2-yearly burnt plots than on control plots(721 versus 1207 kg handash1 for the woodland site and 1497 versus1996 kg handash1 for the grassland site) but treatments were onlystatistically different in some years Herbage mass of perennial



B convergens W 789 lt001 5 (06)a 3 (05)ab 1 (05)b ndash

B convergens G 807 lt001 12 (19)a 4 (10)b 5 (12)ab ndash

C fallax G 518 lt005 159 (187)a 285 (399)b 287 (763)b ndash

E polyphyllus W 116 035 69 (172) 44 (227) 24 (126) ndash


F pauciflora G 460 lt005 11 (15)a 6 (14)ab 5 (29)b ndash

G canescens W 168 022 46 (166) 15 (98) 8 (80) ndash


H contortus W 101 039 743 (1232) 1016 (2395) 1064 (2194) ndash


Table 5 Mean grazing scores (se of mean in parentheses) for the firetreatments in 2007 2011 and 2013 in the grassland plots

Thecontrol treatmentwasnot burnt Early2Early4 andEarly6 treatments areburnt early in the season at 2- 4- and 6-yearly intervals Late 2 Late 4 andLate6 are treatments burnt late in the season at 2- 4- and 6-yearly intervalsDifferent letters denote significantly different means (Plt 005) lsd test

DF= 9 Treatments burnt 2 years previously

Treatment 2007 2011 2013

Control 27 (021)a 13 (007)ab 14 (015)aEarly 2 40 (036)b 18 (052)a 28 (019)bEarly 4 35 (018)b 14 (013)ab 26 (013)bEarly 6 34 (014)ab 13 (024)ab 18 (023)aLate 2 35 (031)b 11 (004)b 27 (005)bLate 4 38 (034)b 14 (035)ab 31 (0003)bLate 6 28 (038)a 11 (008)b 13 (002)a


grasses increased through time on all woodland treatmentsalthough significantly less so for early than late burnt plots(Fig 5c) Herbage mass of perennial grasses was statisticallylower on the early versus late burnt plots in 2013 in the woodlandsite and in 2009 and 2013 in the grassland site (Fig 5d)

Herbage mass of species

In the woodland the herbage mass of B convergens crashedfrom 17 to 4 of total herbage mass (averaged across alltreatments) in 2001 and remained a minor component of thepasture thereafter It was often higher however on burnt plotsand sometimes on 2-yearly burnt plots compared with controls(Appendix 1 Table 6 Fig 7a) Conversely in the grassland sitethe herbage mass of B convergens increased in the later part ofthe experiment although only to very low herbage masses andwas higher on average on early versus late burnt plots (Fig 7b)

The herbage mass of the other major annual grass at thewoodland site E polyphyllus crashed from 22 to 3 of totalherbage mass following the failed wet season of 2002ndash03recovering only slightly on the burnt plots (Fig 7c) There wasa significant year fire frequency interaction with unburntcontrols having among the highest herbage masses ofE polyphyllus at the start of the experiment but by 2011 theherbage mass of E polyphyllus was higher on burnt plots

The herbage mass of Iseilema spp did not recover after thevery dry year of 2003 on any treatment in the grassland site(Fig 7d) Control plots initially had among the highest herbage

masses of Iseilema spp but by 1996 among the lowest(significant year season frequency interaction) so that in thelater part of the study 2-yearly burnt plots often had higherherbage masses of Iseilema spp than the control plots

The herbage mass of the preferred perennial grassesC fallaxand D fecundum declined through the experiment regardlessof fire treatment The herbage mass of H contortus was initiallyless than 250 kg DM handash1 (lt20) at the woodland site but itincreased through the experiment so that by 2007 H contortuswas the dominant pasture component across all fire treatmentsand controls with the contribution to total herbage mass rangingfrom 74ndash85 (Fig 7e) However increases in the per centherbage mass ofH contortus through time were smaller on earlyseason and 2-yearly burned plots (significant year seasonand year frequency interactions) so that by 2011 2-yearlyburnt plots had statistically lower per cent herbage masses ofH contortus than controls and in 2013 late burnt plots had astatistically higher herbage mass ofH contortus than early burntplots

The herbage mass of A latifolia (an unpreferred short-livedperennial grass) increased dramatically after the dry year of 2003in the grassland site across most treatments but the herbage masstended to increase less on burnt plots than on the control plotsalthough the herbage mass of A latifolia was only statisticallylower on burnt plots in 2003 (Fig 7f)

Gomphrena canescens the most prevalent dicot at thewoodland site tended to have a higher herbage mass on burnt

Woo

dy c

over

(

)C

hang

e in

woo

dy c

over

(

)

Fire frequency ()

0

2

4

6

8

10

12

0

0

ndash4

4

8

12

16

20

16 26 47 0 16 26 47

Early Late b Contrburn burn rol

(a) (b)

(c) (d)

Fig 4 Meanwoody cover in (a) woodland and (b) grassland sites in 1995 and change in woody cover in (c) woodlandand (d) grassland sites between 1995 and 2013 in relation to the frequency and season of fires (Bars represent seof mean)


plots and had a significantly higher herbage mass on 2-yearlyburned plots versus control plots in 2005 and 2011 (Fig 7g)Herbagemass of the dominant forb in the grassland the perenniallegume Flemingia pauciflora increased more through time onmore frequently burnt treatments than the control treatment(Fig 7h) and in 2013 was statistically higher on burnt than onunburnt plots

To simplify the interpretation of findings in the study acrossall analytical approaches a summary of understorey variablesrsquoresponse to fire is given in Table 7

Discussion

Treatment effects on fire characteristics

The slightly higher fuel loads in the early comparedwith late firesis due to the loss of herbage to grazing and detachment betweenthe June and October fires Despite the higher fuel loads fires onthe EDS treatments had a lower flame height probably due to thecoolermore humid conditions and less cured fuel This indicatesthatfires on the EDS treatments were less intense than on the LDStreatments consistent with previous studies in northern Australia(Williams et al 1998 Russell-Smith et al 2003 Russell-Smith

and Edwards 2006) The lower fuel loads and flame height on the2-yearly treatments (compared with the 4- and 6-yearly firetreatments) can be explained by removal of the previous yearrsquosstanding dead material by recent fire and the heavier or morefrequent post-fire grazing that the 2-yearly fire treatmentsexperienced The combined effect of higher fuel load and higherflame heights meant that the 4- and 6-yearly fire treatmentsprobably hadmore intense fires during this study consistent withthe general trend of increasing fire intensity with TSF (Murphyand Russell-Smith 2010)

Interactions between fire and grazing

Most experimental fire studies in northernAustralia have been onungrazed land This study was unusual in that the sites were opento grazing and thus the important interactions between fire andgrazing are part of the fire treatments This makes the study morerelevant to grazed systems However the lack of a lsquofire but nograzingrsquo set of treatments means that the different fire treatmentsin this experiment are inextricably linked to post-fire grazinglevels Grazing impacts on the different fire treatments can beinferred through the observed levels of grazing for different

Table 6 Effect of frequency of fire season of fire and any fire treatment compared with the unburnt control treatment on total herbage mass andherbage mass of functional groups and species

Years analysed separately with ANOVA Only years where pair-wise comparison of means (Bonferonni test) are significantly different (Plt 005) are shownC Control E early L late B burnt

Years when fire frequency has a significant effect Years when fire seasonhas a significant effect

Years where fire treatmentsare different to controls

Woodland siteTotal herbage mass 1996C gt2- and 6-yearly 1998C gt2- and 4-yearly 2013 E ltL 1996 1998 B ltCGround cover 1996 2003C gt6-yearly 1998C gt2- and 4-yearly

2011C lt2-yearly2011 E ltL 1996 1998 2003 B ltC

Dicots 2011 2-yearlygtC ndash 2011 B gtCPerennial grasses 2009 2-yearlyltC 2013 2-yearlyltC 2009 E ltL 2013 EltL 1998 2000 2009 2013 BltCAnnual grasses 2000 2- and 4-yearlygtC 2009 2-yearlygtC 2003 E ltL 2005 EltL 2000 2009 2013 B gtC annual grasses 2000 2- and 4-yearlygtC 2005 2-yearlygtC 2013

2-yearlygtC2013 E gtL 1998 2000 2001 2009 2013 B gtC

Brachyachne convergens 2009 2-yearlygtC 2011 2-yearlygtC ndash 1999 2000 2003 2009 BgtCEnneapogon polyphyllus ndash ndash 2011 B gtCGomphrena canescens 2005 2-yearlygtC and 6-yearly 2011 2-yearlygtCHeteropogon contortus ndash 2013 E ltL ndash

H contortus 2009 2-yearlyltC (P= 008) 2011 2-yearlyltC and4-yearly

ndash ndash

Grassland siteTotal herbage mass 1998C and 6-yearly gt2-yearly 2000 4 and 6-yearly

gt2-yearly 2013C gt2- and 4-yearly2013 E ltL 2003 2013 B ltC

Ground cover 1996 E ltL 2007 B ltCDicots 2013 2- and 4-yearlygtC ndash 2013 B gtC dicots 2013 2- and 4-yearlygtC ndash 2009 2013 B gtC legumes 2013 4-yearlygtC ndash 2013 B gtCPerennial grasses 1998 2-yearly lt6-yearly and C 2000 2-yearly lt4 and

6-yearly 2013 2- and 4-yearlyltC2013 E ltL 2003 2013 B ltC

Annual grasses 1998 2-yearlygtC 1999 2-yearlygtC 2009 2-yearlygt6-yearly

1999 E gtL 2000 2005 B gtC

Aristida latifolia ndash ndash 2003C gtBB convergens ndash ndash 2005 B gtCFlemingia pauciflora ndash ndash 2013 B gtCIseilema spp 1998 2-yearlygtC 2005 2- and 6-yearlygtC 2013

2- and 6-yearly gt4 yearly2003 E gtL 2013 EltL 2005 B gtC


treatments although grazing scores were not observed for theearlier part of the study and then only whenmonitoring occurredevery 2 years which was always 2 years post-fire (when post-firegrazing is highest in the first few months to a year post-fire seebelow) Hence it is likely that post-fire grazing was heavier in thefirst year post-fire than recorded here

Grazing animals are attracted to burnt areas (Andrew 1986aDyer et al 2003 Letnic 2004 Vermeire et al 2004) In sub-tropical north-east New SouthWales cattle grazing increased onburnt areas between 3 and 5 months post-fire (Southwell andJarman 1987) although for grey kangaroos this effect lasted upto a year The fire lsquomagnetrsquo effect on cattle grazing was longerlived at Katherine (Northern Territory) in the tropical savannas(Andrew 1986a) where early dry season burnt pastures hadelevated utilisation levels while regrowing from reserves in thefollowing late dry season through to the following early wetseason and often until a later fire drew them to a more recentlyburnt patch (Andrew 1986a) This is not surprising on thesepastures where nutrient intake often limits cattle growth and thenitrogen and phosphorus content ismuch higher on recently burntpatches (Smith 1960 Winter 1987 Bennett et al 2003)

In this study both wallabies and cattle were observed to grazeon burnt areas Elevated grazing was still present 2 years post-fire

in the grassland in the years of lower herbage growth in 2007 and2013 The heavier grazing of early versus late burnt plots inthe year of high herbage growth in 2011 suggests post-firegrazing management may be particularly important for early dryseason burnt sites especially where soil moisture permits someregrowth during the dry season The impact of post-fire grazingwill vary from year to year depending on post-fire soil moisturerainfall and stocking rate effects on utilisation

Woody cover

Fire treatment impacts

While there is ongoing debate about the drivers of vegetationthickening in savannas globally (Sankaran et al 2005 Bond2008) this study has confirmed the moderating influence that firecan have on woody cover (Crowley et al 2009 Smit et al 2010Murphy et al 2014)

Season

The smallest increase in woody cover occurred with the LDStreatmentswhich despite the slightly lower fuel loads had higheraverageflameheights This is in contrast to earlierwork in 1999 atthe sites of this experiment (Dyer 2001) where season of fire did

0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000

Unburnt control Early 2 Late 2 Early 4 Late 4

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)

Fig 5 Mean total herbage mass of (a) woodland and (b) grassland sites and mean herbage mass of perennial grassesof (c) woodland and (d) grassland sites with different fire treatments over the period of the experiment (Bars represent seof mean)


not have a detectable effect on woody cover although lateseason fires did cause greater plant damage than early fires in thegrassland site Smaller increases in woody cover through timewith late season fire is consistent with previous studies (Williamset al 1999) andoccurs becausehotter anddrier prevailingweatherconditions cause higher-intensity fires that have a higher levelof top kill for a given fuel load ground cover and plant height(Dyer 2001)

The impact of season of fire on change in woody cover isnot however always so clear cut For example there was nodifference in change in woody cover between the 2-yearly earlyand late burn treatments in the woodland site In this instanceearly fires were just as successful as late fires when implementedevery second year The lower fuel loads in the 2-yearly latetreatments appear to have offset any advantages from the drierhotter conditions later in the year

Frequency

The lower woody cover or rate of increase in woody coverwith increasing fire frequency found in this study was found atthe sites previously in 1999 (Dyer 2001) and has been reportedelsewhere in Australian (Murphy et al 2014) and Africansavannas (Furley et al 2008 Smit et al 2010 Levick et al 2012)

There were exceptions to this conclusion Despite doubling ofthe fire frequency the 2- and 4-yearly late fire treatments at bothsites had similar changes in woody cover through time Top killmay be reduced because of less fuel load and lower fire intensityon the 2-yearly regime Hence the effect of higher fire frequencyon woody cover with 2-yearly fires may be offset by the lowerfire intensity due to lower fuel loads More frequently elevatedpost-fire grazing on 2-yearly burnt plots could also promotewoody plant growth by reducing competition between the under-and overstorey (eg February et al 2013) This is an example ofwhere more fire did not produce better outcomes for themanagement of woody cover in this grazed landscape Too high afrequency led to less effective fires which provided no advantagein terms of woody cover but was counter-productive for pasturecomposition (see later)

Thewoodland6-yearlyfire treatmentshadnotbeeneffectivelyburnt for 12 years in 2013 which explains the similar increases inwoody cover to the controls Fires are generally less effective onwoody plants taller than 2m at the site (Dyer 2001) and woodyplants can grow to above 2m within 13 years allowing them toescape the severe effects of most fires (Dyer 2001) Hence manywoody plants on the 6-yearly fire treatments may have alreadyescaped the flames by growing above flame height The current

0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

Fig 6 Proportion of total herbage mass of (a) legumes on the grassland site (b) annual grasses on the woodland site and(c) herbagemass of annual grasses on the grassland site for thefire treatments over the periodof the experiment (Bars representthe se of mean)


0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )

Fig 7 Mean herbage mass on the woodland site of (a) Brachyachne convergens (c) Enneapogon polyphyllus(e) Heteropogon contortus and (g) Gomphrena canescens and mean herbage mass on the grassland site of (b) Brachyachneconvergens (d) Iseilema spp (f)Aristida latifolia and (h)Flemingiapauciflora in response to thefire treatmentsover theperiodof the experiment (Bars represent the se of mean)


more wooded structure is probably relatively immune to futurefire and hence likely to persist unless perturbed by drought

Comparison between sites

The lowerwoodycover in thegrassland site comparedwith thewoodland site is consistent with the negative correlation betweenwoody cover and soil clay content characteristic of northernAustralia (Williams et al 1996) The higher absolute and relativeincreases in woody cover in the woodland compared with thegrassland site reflects the different potential for woody cover ofthe sites The higher fuel loads combined with the shorter heightof many of the woody plants in the grassland may also haveresulted in greater top kill and mortality of woody plants with firetreatments in the grassland site (Dyer 2001) further contributingto lower rates of increase in woody cover with fire treatments onthis site

Comparison with other studies

Table 8 compares the change in woody cover in this study toother studies in the region The annual change in woody cover inthe 18-year time frame observed here was 75 and 3 times higher(for the woodland and grassland sites respectively) than themean cover change estimated in a meta-analysis of studiesacross northernAustralia (008of ground area yearndash1) (Murphyet al 2014) However increases in woody cover observed in theVictoria River District on historically grazed alluvial savannadominated by E microtheca and E parviflora (Fensham andFairfax 2003 Sharp andWhittaker 2003) were similar to the less

intensefire treatments in the grassland site (E2 andL6) The lowerrate of increase in woody cover on a Terminalia spp land type(Sharp and Whittaker 2003) and downs land type (Fenshamand Fairfax 2003) was similar to more intense fire treatments inthe grassland site in this study (L4) Rates of change in woodycover on limestone hills in Fensham and Fairfax (2003) werecomparable to only the most severe fire regime (L2) in thewoodland site in this study The declining change inwoody coverobserved on limestone plains in Fensham and Fairfax (2003) wasin stark contrast to the increasingwoody coverwith even themostsevere fire regime (L2) in the woodland site in this study

Implications for regional trends in woody cover

The higher rates of increase in woody cover in this study mayhave been influenced by several factors higher rainfall in theregion since the 1970s (Shi et al 2008) as woody cover potentialincreases with higher rainfall (Williams et al 1996) enhancedatmospheric CO2 concentrations (Buitenwerf et al 2012) andcurrent levels of cattle grazing

The response in woody cover to fire in this study is almostcertainly due to the interacting effects of fire and grazing whichwas often higher on recently burnt plots When herbage isremoved by grazing there is less competition between woodyplants and the herbage layer (Scanlan 2002 February et al 2013)which can advantage woody growth Grazing also decreases fuelloads which lowers fire intensity again leading to higher woodygrowth Consequently even with a similar fire frequency andseasonal timing on grazed versus ungrazed land the severity of

Table 7 Summary of responses of herbage mass of functional groups and species to fire across all analysesns No significant effect detected W Woodland G grassland

Functional groupspecies Site Effect ofmore recentfire (TSF) Analyses across years Analysis of single years

Annual grasses W Decreaser in 1996 Increaserlater

Increaser 2-yearly fire Increaser 2-yearly and fire vs control

Annual grasses G Increaser in later part of study Increaser 2-yearly fire Increaser 2-yearly fireDicots W Increaser Increaser 2-yearly fire Increaser 2-yearly fireDicots G Increaser Increaser 2-yearly and early

fireIncreaser 2- and 4-yearly fire

Perennial grasses W Decreaser Supressed 2-yearly fire Decreaser early firesPerennial grasses G Decreaser Trend decreaser 2-yearly fire

(ns)Decreaser early and 2 yearly

Aristida latifolia G Decreaser ns Decreaser fire vs controlBrachyachne convergens W and G Increaser up to 3 years post-fire Increaser 2-yearlyfire (W)and

early fire (G)Increaser 2-yearly fire and fire vs

controlChrysopogon fallax G Neutral Neutral NeutralDichanthium fecundum W and G Decreaser trend ns Neutral NeutralEnneapogon polyphyllus W Decreaser in annual-

dominated pastureincreaser in perennial-dominated pasture

Decreaser in annual-dominated pastureincreaser in perennial-dominated pasture

Increaser fire vs control

Flemingia pauciflora G Increaser in last year Increaser with more frequentfire


Gomphrena canescens W Increaser in later years of study ns Increaser 2-yearly fire in later part ofstudy

Heteropogon contortus W Decreaser in later years ofstudy

Suppressed early and 2-yearlyfires

Decreaser early fires and 2-yearly fire

Iseilema spp G Inconsistent Increaser 2-yearly early part ofstudy

Increaser 2-yearly fire inconsistentseasonally


fire would be lower on grazed land due to lower fuel loads Lesssevere fires combinedwith the competitive advantage for woodygrowth under grazing could be expected to lead to higher rates ofincrease in woody cover on grazed land

An analysis of fire frequency in the Victoria River Districtbetween 1997 and 2010 (Cowley and Jenner 2013) howeverfound a much lower fire frequency on land used for cattleproduction than on indigenous defence and conservation landThis was particularly so for highly fertile land types where fireintervals averaged just 1 in 10 years on pastoral land comparedwithmore than 1 in 3years in ungrazed landscapesThis studyhasdemonstrated that this lower incidence of fire on grazed land willinevitably lead to higher woody cover

Rainfall enhanced atmospheric CO2 concentrations grazingand the related lower frequency and severity of fires on grazedland may all be interacting to increase the growth rates of woodyplants on grazed land in the region reducing the time required forwoody plants to grow high enough to escape the flames (Bond2008) The unanswered questions are what is the new bioticpotential for woody cover in this region with higher atmosphericCO2 concentrations and rainfall when will it stabilise andwhat if any fire regime will act to preserve the historically moreopen savanna structure or will the new woody potential be aforested system (Higgins andScheiter 2012)Alternatively thereis little certainty that the current higher rainfall will persist underclimate change Declining rainfall in north-western Australia ispredicted when the effect of Asian aerosols (which have beenimplicated in the recent higher rainfall) is removed from climatemodels (Rotstayn et al 2007) Under a scenario of lower rainfallextended droughts could become more frequent promotingdieback events that would reduce woody cover in the region(Fensham et al 2009)

Further analysis of aerial photos between 1995 and 2013 isrecommended to reveal shorter-term responses inwoody cover tofire regime and to establish whether woody cover is still

increasing Whether browsing of woody plants was higher post-fire is unknown as this was not monitored but it may be a usefularea of future research in understanding woody response to fire atthe site as elsewhere the combination of fire and browsing act tosuppress woody cover (Staver et al 2009)

Herbage mass and ground cover

Lower herbagemass and ground cover post-fire is consistent withprevious studies (Dyer et al 2003 Fuhlendorf and Engle 2004Kutt and Woinarski 2007 Russell-Smith et al 2010 Vermeireet al 2014) but was relatively short-lived here consistent withfindings at the site in 1999 (Dyer 2001) (usually only in thefirst year post-fire)

Lower herbagemass andgroundcover in thefirst year post-firecan be explained by the removal of the previous seasonrsquos standingdead herbage mass by fire and increased grazing of post-fireherbage regrowth Reduction in post-fire herbage mass wasgreater in the grassland than thewoodland site whichmay be dueto the higher stocking rates and post-fire grazing on that site

The occasionally lower herbagemass and cover on early burntplots compared with late burnt plots may be reflecting heavierpost-fire grazing on early burnt plots through the removal ofrecent herbage growth or indirectly by leading to poorer pasturecondition and hence lower growth Alternatively the higherwoody cover on early burnt plots could also lead to lower herbagegrowth on early burnt plots

Seasonal or grazing mediated shifts in compositionof herbage

Therewere large compositional changes through time at both sitesthat were unrelated to fire treatment and were more likely to havebeen driven by rainfall or grazing pressure The high stockingrates (above recommended rates for a median growth year) inthe failed wet season of 2003 left a lasting change in speciescomposition at both sites

Table 8 Change in woody cover observed in this study compared with other studies on similar land types in northern AustraliaE2 Early 2-yearly fire L2 late 2-yearly fire L4 late 4-yearly fire L6 late 4-yearly fire

Other northern Australian studies Comparable experimental treatments 1995ndash2013Study Dominant overstorey

vegetation typeWoody cover

change yearndash1Years Treatment Woody cover

change yearndash1

Murphy et al (2014) Mean across reviewednorthern savanna studies

008 1940s tomid-2000s

Average woodlandAverage grassland

060024

Sharp and Whittaker (2003) Eucalyptus microtheca 022 1948ndash93 Grassland E2Grassland L6

017024

Sharp and Whittaker (2003) Excoecaria parviflora 018 1948ndash93 Grassland E2Grassland L6

017024

Sharp and Whittaker (2003) Mixed E microthecaE parviflora

021 1948ndash93 Grassland E2Grassland L6

017024

Sharp and Whittaker (2003) Terminalia spp 006 1948ndash93 Grassland L4 004Fensham and Fairfax (2003)A Alluvia ndash E microtheca 013 1948ndash93 Grassland E2 017Fensham and Fairfax (2003)A Downs ndash Bauhinia

cunninghammigrassland008 1948ndash97 Grassland L4 004

Fensham and Fairfax (2003)A Limestone hills ndash Corymbiaterminalis

017 1949ndash97 Woodland L2 019

Fensham and Fairfax (2003)A Limestone plains ndash Eucalyptustectifica Corymbiaterminalis

ndash001 1948ndash93 No similar cover changeon woodland

ndash

ASum of cover increases for lt3-m and gt3-m size classes


The conversion from an arid short grass-dominatedcommunity toonedominatedby theperennial grassHcontortusin the woodland site is unlikely to be fire-related because changesoccurred on all burnt and unburnt treatments The same shift incomposition has been observed previously inside and outsidegrazing exclosures on the Victoria River Research Station(albeit at a slower rate when grazed) (Foran et al 1985 Bastinet al 2003) and has been interpreted as recovery from the initialdegraded pasture state facilitated by lower grazing levels(Stockwell et al 1994 Bastin et al 2003 Watson and Novelly2012)

The broad geographic range of reported increases inH contortus across north-western Australia implicates increasedrainfall in the region Heteropogon contortus occurs in the700ndash1200-mm rainfall zone (Weston 1988) The average rainfallat the Victoria River Research Station in the 20 years before thestudy was 683mm which is marginally less than ideal for thisspecies Heteropogon contortus was present at the site in 1994but only as a relatively minor component of the pasture Duringthe study the average rainfall was 851mm which is within therainfall range suitable for H contortus Its seed production (Orret al 2004) density and basal area (Orr and OrsquoReagain 2011)respond quickly to wetter seasonal conditions Hence increasedrainfall may have driven the observed increases in the herbagemass and ground cover of H contortus in this study elsewhereon the Victoria River Research Station (Bastin et al 2003) and innorthern Western Australia (Watson and Novelly 2012) Thefaster transition from arid short grass to H contortus seen insidegrazing exclosures (Bastin et al 2003) reflects the sensitivity ofdensity and plant survival of H contortus to grazing (Orr et al2010a 2010b) and suggests grazing levels are likely to havemediated the rate of compositional changes under the currenthigher rainfall regime

The increase in H contortus was mirrored by staggereddecreases in arid short grasses at the woodland site The first aridshort grass to crash was the annual B convergens in 2001followed by the short-lived perennial E polyphyllus whichcrashed following the failed wet season of 2002ndash03 Both thesespecies were replaced by H contortus The concomitant loss ofthe preferred perennial grass D fecundum across the woodlandsite [but not inside exclosures in the same paddock (Bastin et al2003)] is presumed to be due to selective grazing The selectionpressure on D fecundum would have increased as preferredspecies were replaced by the less preferred H contortus Thechange in vegetation state in the woodland site can be consideredboth positive (from shorter- to longer-lived grasses) and negative(less preferred species) from a pastoral perspective but seemslikely to be relatively stable under current ambient stocking ratesand higher rainfall in the region

The decline of the preferred perennial grasses C fallax andD fecundum and the preferred annual grasses Iseilema spp inthe grassland occurred regardless of fire treatment and wasaccompanied by the concomitant increase in the less preferredshort-lived perennial A latifolia and annual B convergensgrasses The combination of very low rainfall and higher-than-recommended stocking rate in 2003 would have resulted in veryhigh utilisation of Iseilema spp C fallax and D fecundumSimilar changes in vegetation composition for these vegetationcommunities have previously been attributed to heavy utilisation

(Stockwell et al 1994) Brachyachne convergens is a knownincreaser under higher grazing (Fisher 2001 Cowley et al 2007)although the herbage mass of A latifolia responds more toseasonal variation than grazing (Phelps 2007) Hence the morevariable rainfall (and hence utilisation levels) in the latter part ofthe study may have contributed to increases in A latifolia andB convergens in the grassland site

Fire-related changes in species

The increases in grazing-avoidant annual grasses and unpreferredforbs and declines in perennial grasses with more frequent orearly fire in this study suggest changes were driven by higherlevels of post-fire grazing

Season offire had less of an effect on species composition thanthe frequency of fire but there were some seasonal responsesEarly burnt plots hadhigher herbagemasses of annual grasses anddicots and a lower herbage mass of perennial grasses in somelater years probably due to the higher grazing levels on earlyburnt plots Thiswas in contrast to earlierfindings at the site (Dyer2001) when season of fire did not affect species compositionpossibly because stocking rates were lower in the earlier part ofthe study and fires were more likely to be implemented in otherparts of the paddock to reduce post-fire grazing pressure on thesites recently burnt

Burning has previously been found to promote H contortusbut more so if livestock were removed or reduced following fire(Orr et al 1991 1997 2010a 2010b Orr and Paton 1997 Furleyet al 2008) Burning and light grazing increases recruitmentand density of H contortus (Orr et al 2010b) enhancing soilseedbanks and germination (Campbell 1996) However heavygrazing reduces the herbage mass of H contortus (Orr et al2010a) both through removal of recent growth by grazing butalso lower growth due to decreased survival density basal areaseedling recruitment and soil seedbanks (Orr et al 2010a 2010b)Heteropogon contortus is more preferred by cattle when itsherbage mass is low (Hendricksen et al 2010) and when it isburnt (Paton and Rickert 1989 Orr et al 1997) Given thetendency for heavier grazing on 2-yearly and early burnt plots(although not found on thewoodland site) heavy post-fire grazingof H contortus may explain its lower increase through time onthese treatments

The three major short-lived grasses E polyphyllus Iseilemaspp and B convergens responded differently to fire through thestudy The disturbance-tolerant short-lived annualB convergenstended to have a higher herbage mass on burnt treatmentsconsistent with previous findings at the site (Dyer 2001) andelsewhere where burning has been found to increase annualgrasses (Russell-Smith et al 2003) Its higher herbage masson early burnt plots in the grassland site in 2005 suggests thatthe higher post-fire grazing here may also be a factor asB convergens is a widely recognised increaser in disturbed areasand under high levels of grazing (Foran et al 1985Milson 2000Fisher 2001 Cowley et al 2007)

In contrast the facultative perennial E polyphyllus wasinitially suppressed by fire at the site while the pasturewas still anarid short grass community This trend reversed as the woodlandsite transformed to a perennial tall grass-dominated communitywhen E polyphyllus then tended to increase on burnt plots


E polyphyllus tends to be dominant at intermediate levels ofpasture condition (Hacker and Tunbridge 1991 Stockwell et al1994 Watson and Novelly 2012) It has been found to increaseunder burning and grazing in central Queensland (Kutt andWoinarski 2007) and to increase at very high and very lowgrazing pressures in northern Mitchell grasslands (Fisher 2001)and with higher stocking rates in the Kimberley (Hacker andTunbridge 1991) The varied response of E polyphyllus to fire asthe composition of the vegetation community changed over timesuggests that the combination of fire and post-fire grazing willtend to push the vegetation community to a poorer condition stateand the resultant composition will vary depending on the pre-existing assemblage When the pasture was a short-lived aridshort grass-dominated community E polyphyllus was the tallerlonger-lived component anddeclinedunderfire at the expense ofincreases in the shorter short-lived B convergens When thepasture was perennial-dominated fire tended to reduce the tallerlonger-lived H contortus but increase the smaller shorter-livedE polyphyllus The herbage mass of Iseilema spp tended toincrease post-fire but only in someyearswhich is consistentwithpreviousfindings at the site (Dyer 2001)Most of these yearswerewhen stocking rateswere quite low at the site (eg 1998 2005 and2013) Hence post-fire grazing may bemasking the fire responseof this preferred genus as it may have been grazed from thepasture before the June sampling (Foran et al 1985) particularlywhen utilisation rates were high

Fire has previously been found to reduce the herbage mass ofA latifolia both with (Orr et al 2010a) and without (Orr et al1991) grazing Cattle are more likely to graze A latifolia plantswhen the herbage mass is low (Phelps 2007) so fire could alsoincrease grazing pressure on this less preferred species post-fireIn Queenslandrsquos Mitchell grasslands grazing reduced theinflorescence density of A latifolia (Phelps 2007) and burningreduced seed production and soil seed banks Therefore fire andenhancedpost-fire grazing could be acting in concert to reduce theherbagemass ofA latifolia onmore frequently and recently burnttreatments in this study

Native forbs and legumes have previously been found toincrease post-fire in tropical Eucalypt savannas (Fensham 1990Williams et al 2003a) Fire may act to break the dormancy of thenative legumes as found elsewhere (Williams et al 2003bWilliams et al 2005) The forbs that increased with fire in thelater part of this study (F pauciflora and G canescens) areunpreferred so may also have been advantaged where there ishigh post-fire grazing

Management implications

To burn or not to burn

Cattle producers are often reluctant to burn becauseimplementing fires requires careful planning skills andmanagement and involves at least a short-term loss of carryingcapacity and livestock production However this study hasdemonstrated that in the longer term a lack of fire will increasewoodycoverwhich is equally likely to reducepasturegrowth andcarrying capacity (Dyer and Stafford Smith 2003) Furthermoreif livestock numbers are not reduced to match reductions incarrying capacity with higher woody cover lower pasturegrowth leads to higher pasture utilisation and a decline in land

condition further reducing carrying capacity and productionpotential

How often to burn

Two-yearlyfireswere unnecessary formanagingwoodycoverand deleterious to pasture composition in this semiarid savannaand the 6-yearly fires were not effective in managing woodycover Four-yearly fires were optimal in this environment formanagingwoodycoverwhilemaintaininghealthypastures Four-yearly fires are more frequent than previous recommendationsfor managing woody cover in the semiarid savannas of theVictoria River District (every 5ndash7 years) suggested byDyer et al(2001) However the 6-yearly burning cycle in this study mayhavebeenmore successful atmanagingwoodycover if it hadbeenimplemented when conditions were better for burning so cannotbe entirely discounted

Early versus late dry season fire

Four-yearly late fires were needed to manage woody coveron both the woodland and grassland sites Early fires were noteffective in managing woody cover and were sometimesassociated with undesirable shifts in species composition tomoreannual grasses and unpreferred forbs presumably due to longerpost-fire grazing impacts However implementing late seasonfires is not without potential hazards as they are more difficult tomanage and are more likely to escape planned burn zones Theyare also less likely to be implemented due to lower fuel loads laterin the year unless stocking rates are adjusted to achieve adequatefuel loads Hence implementing a prescribed late season firerequires significant forward planning and skill

Early fires are sometimes promoted to ensure there is enoughherbage mass to burn (before the grass is eaten by livestock)but wet season spelling from grazing may be a better optionpromoting the dual purpose of recovery of pasture condition andensuring enough herbage mass for a more effective fire later inthe dry season

Burning late (compared with early) in the dry season also hasseveral other advantages for cattle production enterprises Itminimises the length of time between burns and the next rainsand the time of no or low herbage availability which makesmanagement of livestock in burned paddocks less problematic orless expensive if agistment or sale of livestock is required toenable burning However if paddocks need to be spelled for awet season before burning to build up enough fuel (eg from theprevious October) and then are spelled after burning in thefollowing October until the following June paddocks could beout of production for up to 18 months to achieve a fire

Late dry season fire has been implicated in declines in tropicalfauna (Andersen et al 2012 Woinarski and Legge 2013)Burning different paddocks or parts of paddocks each year sothat there are always unburnt parts of the landscape available asrefugia may assist to ameliorate late fire effects on biodiversityalthough leaving some parts of the landscape unburnt for a longtime is also recommended (Andersen et al 2012 Woinarski andLegge 2013)

Management of livestock in burnt paddocks

This study suggests that preferential grazing of burned areascan lead to undesirable changes in species composition as was


found on more frequently and early burned treatments Ideallypaddocks should be spelled immediately after fire until the end ofthe following growing season Where livestock are left inpaddocks when only parts of paddocks are burned burned areasshould be large enough to sustain post-fire grazing of all thelivestock remaining in the paddock and stocking rates should beadjusted to the post-fire available herbage (non-burned area) atthe paddock scale

Impacts of fire on spatial use of the range need to beincorporated into fire planning Fire can be placed in areas withunutilised rank pastures to promote use by livestock Ideallyareas that are already heavily preferred by cattle such as closer towater or preferred pastures should not be burned if cattle are toremain in the paddock as this will compound already existingpreferences and lead to high utilisation of these areas (Dyer et al2003)

Flexibility in fire intervals

This study followed the experimental design as closely asseasons permitted with rigid 2- 4- and 6-yearly intervalsregardless of seasonal conditions which sometimes led toineffective or failed burns In practice landmanagers can bemoreflexible in implementing fires choosing not to burn in poorerseasons when forage and fuel are limiting and fires would beineffective or spelling paddocks to achieve required fuel loadswhen fires aremost effective late in the dry season Alternativelyburning can be implemented in years of high herbage growthwhen fuel loads are higher andfires likely to have greatest impactson woody cover Higher availability of herbage also increasesflexibility in managing livestock numbers pre- and post-firebecause herbage utilisation will generally be lower in years ofhigh herbage growth so that livestock can be dispersed betweenother paddocks to enable spelling of burned paddocks withoutresulting in over-utilisation of the other paddocks (Scanlan et al2011) Where seasonal forecasts have useful predictive power(eg Cobon andToombs 2013) they can be used to plan burns forwhen there is a higher likelihood of receiving at least medianrainfall in the following wet season which will reduce the impactof fire on herbage mass and ground cover in the following yearHowever while some flexibility in fire interval may be useful ifmanagement of woody cover is the goal it is still imperative toensure thatfire intervals are close enough to preventwoody plantsgrowing to above the flame height between fires when theybecome invulnerable to fire

Optimal fire management for pastoral and other land uses

The recommendation for late dry season fires on pastoral landconflicts with recommendations for early dry season fire onindigenous and conservation land across northern Australiawhere frequent hot late dry season fires have been implicated indeclines in biodiversity (Legge et al 2011 Russell-Smith et al2012 Woinarski and Legge 2013) and enhanced greenhouse gasemissions (Russell-Smith et al 2013) The apparent disparityreflects the widely varying current fire regimes on pastoral versusother land uses Compared with indigenous and conservationlandfire frequency isgenerallymuch loweronpastoral land in theVictoria River District (Cowley and Jenner 2013) The mostproductive grasslands in pastoral regions currently burn lessfrequently (1 in 10 years) than the 1 in 4 years that this fire study

suggests is optimal and compared with the higher than 1-in-3-yearfire interval currently ondefence and conservation land in theVictoria River District (Russell-Smith et al 2010 Cowley andJenner 2013) The current low fire frequency on productivepastoral land would allow woody plants to grow above the flameheight between fires and become relatively immune to futureimpacts fromfireModerate productivity red soils onpastoral landin the Victoria River District (equivalent to the woodland site inthis study) are burnt on average 1 in 6 years (Cowley and Jenner2013) which in this study was ineffective in managing woodycover Hence while fire frequency needs to be reduced offthe pastoral estate (Russell-Smith et al 2013) on pastoral landfire frequency often actually needs to be increased Similarlywhile cooler early dry season fires are promoted for biodiversityconservation and reducing greenhouse gas emissionsimplementing this on pastoral land would lead to increasingwoody cover and pasture decline Optimal fire management willvary with prevailing fire regimes and land management goals

The value of long-term research

This studyoffire is unique inAustralian tropical savannas in that itis long-term replicated with controls conducted on woodlandand grassland sites and grazed Its location in the heart of thetropical savanna pastoral zonemakes it particularly relevant to themanagement of northern Australian grazed lands This study hasprovided new insights into the consequences of fire in the grazedcontext and in the context of the broadervegetation changes at thesite thatwere drivenbygrazing and rainfall Thevalueof the studyincreased with its duration Major changes in pasture speciescomposition did not occur until the latter half of the study andwould have beenmissed if the study had been discontinued at theend of the initial experimental period in 1999 (Dyer 2001) ormisinterpreted without the aid of the unburnt control treatmentsdemonstrating the value of long-term controlled studies(Lindenmayer et al 2012) Furthermore questions can now beaskedof this study thatwere originally not envisaged as importantor relevant to the grazing industry 20 years ago such as thelong-term impacts of fire on slow-response variables such asabove- and below-ground carbon storage and plant speciesdiversity This study site may now be useful in global meta-analyses of the drivers of savanna structure To facilitate thefurther use of this dataset it is now freely available to researchersandmodellers globally via open access at httpportalaekosorgaudataset152559 (NTDPIF 2014)

Conclusion

This study has demonstrated the important role of late season firein themanagement ofwoodyvegetation cover in semiarid tropicalsavannas Late dry season burning every 4 years stabilisedwoodycover without deleterious effects on pasture compositionImplementing early dry season fires [as currently recommendedfor biodiversity conservation in northern Australia see Russell-Smith et al (2013) and Skroblin et al (2014)] on moderate tohighproductivity pastoral landwould lead to continuing increasesin woody cover and declines in carrying capacity for livestockand pasture condition Indeed the inappropriate application offire regimes on pastoral land could ultimately contribute to landdegradation and loss of pastoral productivity of some of the most


productive land types across northern Australia This study hascontributed to the conclusion that optimal fire management mayvary depending on vegetation community land use and thefrequency and seasonal timing of fires

Acknowledgements

Thanks are due to Whitney Dollemore Kieren McCosker and DionneWalshfor provision of stocking rate data Thanks also to the many people who haveassisted with data collection at the site over the years and Stefan Maier andJeremy Russell Smith for advice regarding the aerial woody cover analysisThe authors thank Dr Gabriel Crowley and two anonymous reviewers fortheir constructive comments on the manuscript This project was funded byNT Department of Primary Industry and Fisheries and by Meat LivestockAustralia until 2001

References

Andersen A N Woinarski J C and Parr C L (2012) Savanna burningfor biodiversity fire management for faunal conservation in Australiantropical savannas Austral Ecology 37 658ndash667 doi101111j1442-9993201102334x

Andrew M H (1986a) Use of fire for spelling monsoon tallgrass pasturegrazed by cattle Tropical Grasslands 20 69ndash78

Andrew M H (1986b) Selection of plant species by cattle grazing nativemonsoon tallgrass pasture at Katherine NT Tropical Grasslands 20120ndash127

Archer S (1995) Herbivore mediation of grass-woody plant interactionsTropical Grasslands 29 218ndash235

Bastin G Ludwig J Eager R Liedloff A Andison R and Cobiac M(2003) Vegetation changes in a semiarid tropical savanna northernAustralia 1973ndash2002 The Rangeland Journal 25 3ndash19 doi101071RJ03001

Bennett L T Judd T S and Adams M A (2003) Growth and nutrientcontent of perennial grasslands following burning in semi-arid sub-tropical Australia Plant Ecology 164 185ndash199 doi101023A1021253600712

BondW J (2008)What limits trees in C4 grasslands and savannas AnnualReview of Ecology Evolution and Systematics 39 641ndash659 doi101146annurevecolsys39110707173411

BondW J andKeeley J E (2005) Fire as a global lsquoherbivorersquo the ecologyand evolution of flammable ecosystems Trends in Ecology amp Evolution20 387ndash394 doi101016jtree200504025

Bond W J and Midgley G F (2000) A proposed CO2-controlledmechanism of woody plant invasion in grasslands and savannas GlobalChange Biology 6 865ndash869 doi101046j1365-2486200000365x

BondW J andMidgley J J (2001) Ecology of sprouting in woody plantsthe persistence niche Trends in Ecology amp Evolution 16 45ndash51doi101016S0169-5347(00)02033-4

Briske D (1998) Strategies of plant survival in grazed systems a functionalinterpretation In lsquoThe Ecology and Management of Grazing Systemsrsquo(Eds J Hodgson and A W Illius) pp 37ndash67 (CAB InternationalWallingford UK)

Buitenwerf R Bond W J Stevens N and Trollope W S W (2012)Increased tree densities in South African savannas gt50 years of datasuggests CO2 as a driver Global Change Biology 18 675ndash684doi101111j1365-2486201102561x

Burrows W H Henry B K Back P V Hoffmann M B Tait L JAnderson E R Menke N Danaher T Carter J O and McKeonG M (2002) Growth and carbon stock change in eucalypt woodlands innortheast Australia ecological and greenhouse sink implicationsGlobalChange Biology 8 769ndash784 doi101046j1365-2486200200515x

Campbell S D (1996) Plantmechanisms that influence the balance betweenHeteropogon contortus and Aristida ramosa in spring burnt pasturesPhD Thesis The University of Queensland Australia

Cobiac M D (2001) Predicting native pasture growth in the VictoriaRiver District of the Northern Territory PhD Thesis The University ofAdelaide SA Australia

Cobon D H and Toombs N R (2013) Forecasting rainfall based on theSouthern Oscillation Index phases at longer lead-times in Australia TheRangeland Journal 35 373ndash383 doi101071RJ12105

Cowley R A and Bryce D (2004) lsquoKidman Springs Carrying Capacityestimation using pasture growth models and estimates of animal intakersquoNT Department of Primary Industry and Fisheries internal report(DPIF Northern Territory Darwin NT)

Cowley R A and Jenner D (2013) Conservation and indigenous landburns too much and pastoral land not enough in the Victoria RiverDistrict but it depends on land type In lsquoKatherineRuralReviewrsquo Edition316 p 6 December 2013 (Ed J Ward) (Department of PrimaryIndustry and Fisheries Technical Publication Darwin NT) Available atwwwntgovaudContentFilepNL316_14_krrpdf (accessed 13 July2014)

Cowley R A McCosker K D MacDonald R N and Hearnden M N(2007) Optimal pasture utilisation rates for sustainable cattle productionwith a commercial Brahman herd in the Victoria River Downs region ofthe Northern Territory In lsquoNorthern Beef Research Update Conferencersquo(Eds B Pattie and B Restall) pp 34ndash44 (North Australia Beef ResearchCouncil Townsville Qld)

CrowleyGM andGarnettST (1998)Vegetationchange in thegrasslandsand grassy woodlands of east-central Cape York Peninsula AustraliaPacific Conservation Biology 4 132ndash148

Crowley G M Garnett S T and Shephard S (2009) Impact of storm-burning on Melaleuca viridiflora invasion of grasslands and grassywoodlands on Cape York Peninsula Australia Austral Ecology 34196ndash209 doi101111j1442-9993200801921x

Dyer R M (2001) Fire and vegetation management in pasture lands of theVictoriaRiverDistrictNorthernTerritoryMScThesis TheUniversity ofQueensland Brisbane Australia

Dyer R M and Stafford Smith M (2003) Ecological and economicassessment of prescribed burning impacts in semi-arid pastoral lands ofnorthern Australia International Journal of Wildland Fire 12 403ndash413doi101071WF03026

Dyer R M Jacklyn P Partridge I Russell-Smith J and WilliamsD (2001) lsquoSavanna Burning Understanding and Using Fire in NorthernAustraliarsquo (Tropical Savannas CRC Darwin NT)

Dyer R M Cafe L M Cobiac M D and Cowley R A (2003)lsquoDeveloping sustainable grazing management systems for the semi-aridtropics of the Northern Territoryrsquo Final Report to Meat and LivestockAustralia (Meat and Livestock Australia North Sydney NSW)

February E C Higgins S I Bond W J and Swemmer L (2013)Influence of competition and rainfall manipulation on the growthresponses of savanna trees and grasses Ecology 94 1155ndash1164doi10189012-05401

Fensham R J (1990) Interactive effects of fire frequency and site factors intropical Eucalyptus forest Australian Journal of Ecology 15 255ndash266

Fensham R J and Fairfax R J (2003) Assessing woody vegetation coverchange in north-west Australian savanna using aerial photographyInternational Journal of Wildland Fire 12 359ndash367 doi101071WF03022

Fensham R J Fairfax R J and Archer S R (2005) Rainfall land use andwoody vegetation cover change in semi-arid Australian savanna Journalof Ecology 93 596ndash606 doi101111j1365-2745200500998x

Fensham R J Fairfax R J and Ward D P (2009) Drought-induced treedeath in savanna Global Change Biology 15 380ndash387 doi101111j1365-2486200801718x

Fisher A (2001) Biogeography and Conservation of Mitchell Grasslands ofNorthern Australia PhD Thesis Northern Territory University DarwinAustralia

Foran B D Bastin G and Hill B (1985) The pasture dynamics andmanagement of two rangeland communities in the Victoria River District


of the Northern Territory The Rangeland Journal 7 107ndash113doi101071RJ9850107

Fuhlendorf S D and Engle D M (2004) Application of the firendashgrazinginteraction to restore a shifting mosaic on tallgrass prairie Journal ofApplied Ecology 41 604ndash614 doi101111j0021-8901200400937x

Furley P A Rees R M Ryan C M and Saiz G (2008) Savannaburning and the assessment of long-term fire experiments with particularreference to Zimbabwe Progress in Physical Geography 32 611ndash634doi1011770309133308101383

Hacker R B and Tunbridge S B (1991) Grazing managementstrategies for reseeded rangelands in the East Kimberley region ofWestern Australia The Rangeland Journal 13 14ndash35 doi101071RJ9910014

HendricksenR EMylesD J ReidD J andOrrDM (2010) Impacts ofgrazing management options on pasture and animal productivity in aHeteropogon contortus (black speargrass) pasture in central Queensland3 Diet composition in autumn Animal Production Science 50 276ndash283doi101071AN09090

Higgins S I and Scheiter S (2012) Atmospheric CO2 forces abruptvegetation shifts locally but not globally Nature 488 209ndash212doi101038nature11238

Higgins S I Bond W J February E C Bronn A Euston-BrownD I W Enslin B Govender N Rademan L OrsquoRegan S PotgieterA L F Scheiter S Sowry R Trollope L and Trollope W S W(2007) Effects of four decades of fire manipulation on woody vegetationstructure in savanna Ecology 88 1119ndash1125 doi10189006-1664

Jackson J and Ash A J (1998) Tree-grass relationships in open eucalyptwoodlands of north-eastern Australia influence of trees on pastureproductivity foragequality andspeciesdistributionAgroforestrySystems40 159ndash176 doi101023A1006067110870

Jackson J and Ash A J (2001) The role of trees in enhancing soil nutrientavailability for native perennial grasses in open eucalypt woodlands ofnorth-east Queensland Australian Journal of Agricultural Research 52377ndash386 doi101071AR00012

Kgope B S Bond W J and Midgley G F (2010) Growth responses ofAfrican savanna trees implicate atmospheric CO2 as a driver of past andcurrent changes in savanna tree cover Austral Ecology 35 451ndash463doi101111j1442-9993200902046x

Kutt A S and Woinarski J C Z (2007) The effects of grazing and fire onvegetation and the vertebrate assemblage in a tropical savanna woodlandin north-eastern Australia Journal of Tropical Ecology 23 95ndash106doi101017S0266467406003579

Landsberg J Lavorel S and Stol J (1999) Grazing response groupsamong understorey plants in arid rangelands Journal of VegetationScience 10 683ndash696 doi1023073237083

Landsberg J Gillieson D and Salt D (2011) Trees in savanna landscapesAvailable at wwwfirescapecomau (accessed 22 August 2014)

Legge S Murphy S Kingswood R Maher B and Swan D (2011)EcoFire restoring the biodiversity values of the Kimberley region bymanaging fire Ecological Management amp Restoration 12 84ndash92doi101111j1442-8903201100595x

Letnic M (2004) Cattle grazing in a hummock grassland regeneratingafter fire the short-term effects of cattle exclusion on vegetation in south-western Queensland The Rangeland Journal 26 34ndash48 doi101071RJ04003

Levick S R Asner G P and Smit I P J (2012) Spatial patterns in theeffects of fire on savanna vegetation three-dimensional structureEcological Applications 22 2110ndash2121 doi10189012-01781

Lewis D (2002) lsquoSlower Than the Eye Can See Environmental Change inNorthern Australiarsquos Cattle Lands a Case Study from the Victoria RiverDistrict Northern Territoryrsquo (Tropical Savannas CRC Darwin NT)

Lewis D Webb R Boyer D and Turner R (2010) Cattle repeatphotography and changing vegetation in the Victoria River DistrictNorthern Territory Australia In lsquoRepeat Photography Methods and

Applications in Natural Sciencesrsquo (Eds R H Webb D E Boyer andR M Turner) pp 197ndash210 (Island Press Washington DC)

Lindenmayer D B Likens G E Andersen A Bowman D Bull C MBurns E Dickman C R Hoffmann A A Keith D A Liddell M JLowe A J Metcalfe D J Phinn S R Russell-Smith J Thurgate NandWardle G M (2012) Value of long-term ecological studies AustralEcology 37 745ndash757 doi101111j1442-9993201102351x

Ludwig J and Tongway D (2002) Clearing savannas for use as rangelandsin Queensland altered landscapes and water-erosion processes TheRangeland Journal 24 83ndash95 doi101071RJ02004

McAlpine C A Fensham R J and Temple-Smith D E (2002)Biodiversity conservation and vegetation clearing in Queenslandprinciples and thresholds The Rangeland Journal 24 36ndash55doi101071RJ02002

McIvor J G and Gardener C J (1995) Pasture management in semi-aridtropical woodlands effects on herbage yields and botanical compositionAnimal Production Science 35 705ndash715 doi101071EA9950705

Mills A J Rogers K H Stalmans M and Witkowski E T F (2006) Aframework for exploring the determinants of savanna and grasslanddistribution Bioscience 56 579ndash589 doi1016410006-3568(2006)56[579AFFETD]20CO2

Milson J (2000) lsquoPasture Plants of North-west Queenslandrsquo (The Stateof Queensland Department of Primary Industries Brisbane Qld)

Murphy B P and Russell-Smith J (2010) Fire severity in a northernAustralian savanna landscape the importance of time since previous fireInternational Journal ofWildlandFire 19 46ndash51 doi101071WF08202

MurphyB PRussell-Smith J andPrior LD (2010) Frequentfires reducetree growth in northern Australian savannas implications for treedemography and carbon sequestration Global Change Biology 16331ndash343 doi101111j1365-2486200901933x

MurphyBPLehmannCERRussell-Smith JLawesM J andParrK(2014) Fire regimes and woody biomass dynamics in Australiansavannas Journal of Biogeography 41 133ndash144 doi101111jbi12204

Northern Territory Department of Primary Industry and Fisheries (NTDPIF)(2014) Kidman Springs fire experiment 1993ndash2013 Version 1 Obtainedfrom Australian Ecological Knowledge and Observation System DataPortal (AEligKOS wwwportalaekosorgau) made available by TheNorthern Territory Government represented by the Northern TerritoryDepartment of Primary Industry andFisheries (accessed11August 2014)doi1042270553E811AF9BDCF

Orr D M and OrsquoReagain P J (2011) Managing for rainfall variabilityimpacts of grazing strategies on perennial grass dynamics in a dry tropicalsavanna The Rangeland Journal 33 209ndash220 doi101071RJ11032

Orr DM and Paton C J (1997) Using fire to manage species compositionin Heteropogon contortus (black speargrass) pastures 2 Enhancing theeffects of fire with grazing management Australian Journal ofAgricultural Research 48 803ndash810 doi101071A96131

OrrDMMcKeonGM andDayKA (1991)Burningandexclosure canrehabilitate degraded black speargrass (Heteropogon contortus) pasturesTropical Grasslands 25 333ndash336

Orr D M Paton C J and Lisle A T (1997) Using fire to manage speciescomposition in Heteropogon contortus (black speargrass) pastures 1Burning regimes Australian Journal of Agricultural Research 48795ndash802 doi101071A96130

Orr D M Paton C J and Playford C (2004) Dynamics of plantpopulations in Heteropogon contortus (black speargrass) pastures on agranite landscape in southern Queensland 2 Seed production and soilseed banks of H contortus Tropical Grasslands 38 31ndash41

Orr D M Burrows W H Hendricksen R E Clem R L Back P VRutherford M T Myles D J and Conway M J (2010a) Impacts ofgrazing management options on pasture and animal productivity in aHeteropogon contortus (black speargrass) pasture in central Queensland1 Pasture yield and composition Crop amp Pasture Science 61 170ndash181doi101071CP09193


Orr D M Yee M C Rutherford M T and Paton C J (2010b) Impactsof grazing management options on pasture and animal productivity in aHeteropogon contortus (black speargrass) pasture in central Queensland2 Population dynamics of Heteropogon contortus and Stylosanthesscabra cv Seca Crop amp Pasture Science 61 255ndash267 doi101071CP09194

Paton C J and Rickert K G (1989) Burning then resting reduceswiregrass (Aristida spp) in black speargrass pastures TropicalGrasslands 23 211ndash218

Phelps D G (2007) Controlling Aristida latifolia (feathertop wiregrass) inAstrebla spp (Mitchell grass) grasslands with fire and grazing PhDThesis University of New England Armidale NSW Australia

Pressland A (1982) Fire in themanagement of grazing lands in QueenslandTropical Grasslands 16 104ndash112

Rolfe J (2002) Economics of vegetation clearing in Queensland TheRangeland Journal 24 152ndash169 doi101071RJ02008

Rotstayn L D Cai W Dix M R Farquhar G D Feng Y Ginoux PHerzogM Ito A Penner J E Roderick M L andWangM (2007)HaveAustralian rainfall and cloudiness increaseddue to the remote effectsof Asian anthropogenic aerosols Journal of Geophysical Research DAtmospheres 112 D09202 doi1010292006JD007712

Russell-Smith J and Edwards A C (2006) Seasonality and fire severityin savanna landscapes of monsoonal northern Australia InternationalJournal of Wildland Fire 15 541ndash550 doi101071WF05111

Russell-Smith J Whitehead P J Cook G D and Hoare J L (2003)Response of eucalyptus-dominated savanna to frequent fires lessonsfrom Munmarlary 1973ndash1996 Ecological Monographs 73 349ndash375doi10189001-4021

Russell-Smith J Yates C P Brock C and Westcott V C (2010) Fireregimes and interval-sensitive vegetation in semi-arid Gregory NationalPark northern Australia Australian Journal of Botany 58 300ndash317

Russell-Smith J Edwards A C and Price O F (2012) Simplifying thesavanna the trajectory of fire-sensitive vegetation mosaics in northernAustralia Journal of Biogeography 39 1303ndash1317 doi101111j1365-2699201202679x

Russell-Smith J Cook G D Cooke P M Edwards A C Lendrum MMeyer C P andWhitehead P J (2013)Managing fire regimes in northAustralian savannas applying Aboriginal approaches to contemporaryglobal problems Frontiers in Ecology and the Environment 11 e55ndashe63doi101890120251

SankaranMHananN P ScholesR J Ratnam J AugustineD J CadeB S Gignoux J Higgins S I Le Roux X Ludwig F Ardo JBanyikwa F Bronn A Bucini G Caylor K K Coughenour M BDiouf A Ekaya W Feral C J February E C Frost P G HHiernaux PHrabarHMetzgerKL PrinsHHTRingrose S SeaW Tews J Worden J and Zambatis N (2005) Determinants ofwoody cover in African savannas Nature 438 846ndash849 doi101038nature04070

Sankaran M Augustine D J and Ratnam J (2013) Native ungulates ofdiverse body sizes collectively regulate long-term woody plantdemographyand structureof a semi-arid savanna Journal ofEcology1011389ndash1399 doi1011111365-274512147

Scanlan J C (2002) Some aspects of tree-grass dynamics in Queenslandrsquosgrazing lands The Rangeland Journal 24 56ndash82 doi101071RJ02003

Scanlan J C Whish G L Pahl L I Cowley R A and MacLeodN D (2011) Assessing the impact of pasture resting on pasture conditionin the extensive grazing lands of northern Australia In lsquoMODSIM201119th International Congress on Modelling and Simulationrsquo Perth(Eds F Chan D Marinova and R Anderssen) pp 877ndash883 (ModellingSociety of Australia and New Zealand Inc Perth WA)

Scholes R J and Archer S R (1997) Tree-grass interactions in savannasAnnual Review of Ecology and Systematics 28 517ndash544 doi101146annurevecolsys281517

Sharp B R and Whittaker R J (2003) The irreversible cattle-driventransformation of a seasonally flooded Australian savanna Journal ofBiogeography 30 783ndash802 doi101046j1365-2699200300840x

Shi G Cai W Cowan T Ribbe J Rotstayn L and Dix M (2008)Variability and trend of North-west Australia rainfall observations andcoupled climate modeling Journal of Climate 21 2938ndash2959doi1011752007JCLI19081

SkroblinALeggeSWebbT andHuntL (2014)EcoFire regional scaleprescribed burning improves the annual carrying capacity of livestock onpastoral properties by reducing pasture loss fromwildfire The RangelandJournal 36 133ndash142 doi101071RJ13095

Smit I P JAsnerGPGovenderNKennedy-BowdoinTKnappDEand Jacobson J (2010) Effects of fire on woody vegetation structure inAfrican savanna Ecological Applications 20 1865ndash1875 doi10189009-09291

SmithE L (1960) Effects of burning and clipping at various times during thewet season on tropical tall grass range in northern Australia Journal ofRange Management 13 197ndash203 doi1023073894952

SouthwellC J and JarmanP J (1987)Macropodstudies atWallabyCreek3 The effect of fire on pasture utilization by macropodids and cattleWildlife Research 14 117ndash124 doi101071WR9870117

Staver C A and Bond W J (2014) Is there a lsquobrowse traprsquo Dynamicsof herbivore impacts on trees and grasses in an African savanna Journalof Ecology 102 595ndash602 doi1011111365-274512230

Staver A C Bond W J Stock W D van Rensburg S J and WaldramM S (2009) Browsing and fire interact to suppress tree density in anAfrican savanna Ecological Applications 19 1909ndash1919 doi10189008-19071

Stewart G Perry R Paterson S Traves D and Slatyer R (1970) lsquoLandsof the Ord-Victoria Area Western Australia and Northern Territoryrsquo(CSIRO Publishing Melbourne)

Stockwell TGHAndisonRTAshA J Bellamy JA andDyerRM(1994) State and transition models for rangelands 9 Development ofstate and transition models for pastoral management of the golden beardgrass and limestone grass pasture lands of NW Australia TropicalGrasslands 28 260ndash265

Tothill J C (1971)A reviewoffire in themanagement of native pasturewithparticular reference to north-eastern Australia Tropical Grasslands 51ndash10

Tothill J C Hargreaves J and Jones R M (1978) BOTANAL ndash acomprehensive sampling and computing procedure for estimating pastureyield and composition 1 Field sampling Tropical Agronomy TechnicalMemorandumCSIRODivisionofTropicalCrops andPasturesBrisbaneAustralia

Vermeire L T Mitchell R B Fuhlendorf S D and Gillen R L (2004)Patch burning effects on grazing distribution Rangeland Ecology andManagement 57 248ndash252 doi1021111551-5028(2004)057[0248PBEOGD]20CO2

Vermeire L T Crowder J L and Wester D B (2014) Semi-aridrangeland is resilient to summer fire and post-fire grazing utilizationRangeland Ecology and Management 67 52ndash60 doi102111REM-D-13-000071

Walsh D Russell-Smith J and Cowley R (2014) Fire and carbonmanagement in a diversified rangelands economy research policy andimplementation challenges for northern Australia The RangelandJournal 36 313ndash322

Watson I W and Novelly P E (2012) Transitions across thresholds ofvegetation states in the grazed rangelands of Western Australia TheRangeland Journal 34 231ndash238 doi101071RJ11073

Weston E J (1988) Native pasture communities In lsquoNative Pastures inQueensland theResources andTheirManagementrsquo (EdsWHBurrowsJ C Scanlan M T Rutherford) pp 21ndash33 (Queensland Departmentof Primary Industries Brisbane Qld)


Williams R J Duff G A Bowman D M J S and Cook G D (1996)Variation in the composition and structure of tropical savannas as afunction of rainfall and soil texture along a large-scale climatic gradientin the Northern Territory Australia Journal of Biogeography 23747ndash756 doi101111j1365-26991996tb00036x

Williams R J Gill A M andMoore P H R (1998) Seasonal changes infire behaviour in a tropical savanna in northern Australia InternationalJournal of Wildland Fire 8 227ndash239 doi101071WF9980227

Williams R J Cook G D Gill A M and Moore P H R (1999) Fireregime fire intensity and tree survival in a tropical savanna in northernAustralia Australian Journal of Ecology 24 50ndash59 doi101046j1442-9993199900946x

Williams PRCongdonRAGriceAC andClarke P J (2003a) Effectof fire regime on plant abundance in a tropical eucalypt savanna of north-eastern Australia Austral Ecology 28 327ndash338 doi101046j1442-9993200301292x

Williams P R Congdon R A Grice A C and Clarke P J (2003b)Fire-related cues break seed dormancy of six legumes of tropical eucalyptsavannas in north-eastern Australia Austral Ecology 28 507ndash514doi101046j1442-9993200301307x

Williams P R Congdon R A Grice A C and Clarke P J (2005)Germinable soil seed banks in a tropical savanna seasonal dynamicsand effects of fire Austral Ecology 30 79ndash90 doi101111j1442-9993200401426x

Winter W H (1987) Using fire and supplementation to improve cattleproduction from monsoon tallgrass pastures Tropical Grasslands 2171ndash81

Woinarski J C Z and Legge S (2013) The impacts of fire on birds inAustraliarsquos tropical savannasEmu113 319ndash352 doi101071MU12109

Appendix 1 ANOVA of effects of season and frequency of fire year and interactions on total herbage mass ground cover and herbage massof functional groups and species for the woodland and grassland sites 1994ndash2013

Only major species or species with significant fire-related trends are shown

Season(df = 1)

Frequency(df = 3)

Year(df = 13)

Year season(df = 13)

Year frequency(df = 39)

F-value P F-value P F-value P F-value P F-value P

WoodlandTotal herbage mass 178 022 167 025 76 lt0001 175 006 142 008Ground cover 105 033 081 052 85 lt0001 124 027 209 lt001Dicots 043 053 093 047 37 lt0001 114 033 161 lt005 dicots 100 034 099 044 19 lt0001 109 038 127 017Perennial grasses 030 059 151 028 39 lt0001 271 lt001 127 017 perennial grass 006 081 160 026 81 lt0001 129 023 148 006Annual grasses 004 084 237 015 231 lt0001 242 lt001 192 lt001 annual grass 001 093 342 007 151 lt0001 203 lt005 201 lt001Brachyachne convergens 112 032 216 017 62 lt0001 076 065 228 lt001 B convergens 195 020 286 010 46 lt0001 172 009 303 lt0001Dichanthium fecundum 014 071 031 082 12 lt0001 206 lt005 104 043Enneapogon polyphyllus 064 044 074 055 95 lt0001 118 032 211 lt001 E polyphyllus 077 041 037 078 80 lt0001 081 060 299 lt0001Gomphrena canescens 085 038 128 034 19 lt0001 075 071 093 059Heteropogon contortus 015 070 112 040 41 lt0001 183 lt005 093 059 H contortus 001 093 095 046 151 lt0001 257 lt001 224 lt001

GrasslandTotal herbage mass 331 011 181 022 75 lt0001 289 lt001 354 lt0001Ground cover 024 064 011 094 101 lt0001 254 lt001 198 lt001Dicots 004 086 026 085 58 lt0001 123 027 110 035 dicots 015 071 054 067 53 lt0001 248 lt001 243 lt0001 legumes 044 052 077 054 30 lt0001 136 019 229 lt0001Perennial grasses 188 021 235 015 59 lt0001 258 lt001 459 lt0001 perennial grass 030 060 188 021 14 lt0001 065 081 188 lt001Annual grasses 152 025 630 lt005 101 lt0001 108 039 140 009 annual grass 099 035 349 007 35 lt0001 060 085 178 lt001Aristida latifolia 000 098 100 044 25 lt0001 100 044 130 020Brachyachne convergens 895 lt005 101 043 29 lt0001 099 046 082 075Chrysopogon fallax 000 098 020 089 89 lt0001 027 096 185 lt005Dichanthium fecundum 007 079 041 075 22 lt0001 111 036 131 014Flemingia pauciflora 008 078 042 074 22 lt0001 046 093 167 lt005 F pauciflora 039 054 068 059 21 lt0001 156 012 280 lt0001Iseilema spp 019 067 283 011 196 lt0001 512 lt0001 252 lt0001 Iseilema spp 044 053 287 010 87 lt0001 132 025 287 lt0001


wwwpublishcsiroaujournalstrj

has more typically been stable or declined (Russell-Smith et al2010 2012 Murphy et al 2014) Contrasting trajectories ofwoody cover with different fire and grazing management havebeen evident in the Victoria River District of the NorthernTerritory Here some grazed landscapes have undergonewoodland thickening since settlement (Sharp and Whittaker2003 Lewis et al 2010) while in other areas woody cover hasbeen relatively stable or even declined (Fensham and Fairfax2003 Lewis et al 2010) Although both grazing and reducedfrequency of fire on productive pastoral land types have beenimplicated in woodland thickening in the Victoria River District(Lewis 2002 Sharp andWhittaker 2003) potential woody coverin the regionmay also be rising in response to the changing abioticdrivers of increasing rainfall since the 1970s (Shi et al 2008) andrising atmospheric CO2 concentrations

Woodland thickening reduces pasture growth (McIvor andGardener 1995 Jackson and Ash 1998 Scanlan 2002) andsubsequent carrying capacity for livestock production (Dyerand Stafford Smith 2003) although the nutritive value ofpasture is often enhanced under trees (Jackson and Ash 2001Scanlan 2002)

Fire is the only economically viable tool available to managewoody cover (Rolfe 2002 Landsberg et al 2011) and ispreferable to clearing which has adverse impacts on biodiversity(McAlpine et al 2002) ecosystem function (Ludwig andTongway 2002) and greenhouse gas emissions (Burrows et al2002) Fire is also used extensively for other purposes on pastoralproperties as reviewed elsewhere (Tothill 1971 Pressland 1982Dyer et al 2001 2003)

Reducing grazing pressure to allow pasture species tooutcompetewoodyplants or to limit their rate of increase is oftenineffective (Archer 1995) While browsing of woody plants canrestrict their growth (Scholes and Archer 1997 Sankaran et al2013 Staver and Bond 2014) plants prone to woody thickeningtend to be unpreferred (Crowley et al 2009) and in Africansavannas the combination of browsing and fire is required tosuppress tree density (Staver et al 2009)

Fire decreases rates of woody growth (Murphy et al 2010)and stem survival (Williams et al 1999 Murphy et al 2014)However the adaptive traits to fire such as resprouting confervery lowmortality onwoodyplants evenunder regimes of intensefire (Bond and Midgley 2001 Dyer 2001) Hence fire tends toreduce the size and cover ofwoody plants temporarily rather thanaffect their density (Higgins et al 2007)

Fire can also influence composition of the understoreyFire consumes aboveground biomass of herbaceous speciesindiscriminately (Bond and Keeley 2005) In the absence ofgrazing this can provide a competitive advantage to grazing-sensitive species because their less-preferred neighbours are alsoconsumed by fire Burnt pasture however attracts herbivores(Andrew 1986a Southwell and Jarman 1987 Vermeire et al2004) and if burnt patches are relatively small they canbe subjectto very high grazing after fire Hence fire can indirectly promotespecies that avoid grazing through shorter lifespan (temporalavoiders ndash annuals) prostrate or short habit (spatial avoiders) orlow palatability (biochemical avoiders) (Briske 1998 Landsberget al 1999) at the expense of perennial tussock grass speciesthat are more susceptible to grazing Fire can also act to increasegrazing on less-preferred species afterfire (Dyer et al 2001)when

cattle are less selective about the species that they graze (Andrew1986b)

In the light of recommendations for a shift from late to earlydryseason fire to abate carbon emissions and enhance biodiversityconservation (Russell-Smith et al 2013 Walsh et al 2014) thisstudy seeks to improve understanding of the impacts of fireregimes in grazed savannas with a focus on impacts on woodycover and pasture condition Its 20-year duration allowsassessment of the interactions with rainfall and grazing

Methods

Study site

This studywas established in 1993on theVictoriaRiverResearchStation also known as lsquoKidman Springsrsquo 400 km south ofDarwin Northern Territory Australia in the semiarid tropicalsavannas The site has a July to June median rainfall of 742mm(wwwbomgovau rainfall station No 14847 accessed 23August 2014)Consistentwith the increase in the rainfall of north-western Australia since the 1970s (Shi et al 2008) the medianrainfall during the study of 770mm exceeded that of the previous21 years (640mm) Between-year variability in rainfall wasgreater in the studyrsquos last 10 years than during the first 10 years

Twoexperimental siteswere established4 kmapart in separategrazed (by Bos indicus cattle) paddocks of native pasturesin woodland and grassland (Fig 1 Table 1) At each site 16experimental plots (~26 ha in area) were arranged in a 4 4 gridwith each plot separated by cleared firebreaks Two plots were

(a)

(b)

Fig 1 Aerial photograph of the (a) woodland and (b) grassland sites inJune 2013




Data collection







Woodland Grassland
















Time since fire





Results

Site vegetation



Stocking rates





Time since fire

Woody cover





0

1000

2000

3000

4000

0

20

40

60

80

100

0

20

40

60

80

0

20

40

60

Woodland

Grassland

Woodland

Grassland




Aristida latifolia


Her

bage

mas

s (k

g D

M h

andash1)

Gro

und

cove

r (

)

of h

erba

ge m

ass

o

f her

bage

mas

s

(a)

(b)

(c)

(d)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013









Grazing


Woody cover



0

5

10

15

20

25

Sto

ckin

g ra

te (

AE

km

ndash2)

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013























G 060 063 2641 (1060) 2846 (2619) 2803 (1779) 2553 (1222)Ground cover () W 216 015 79 (22) 78 (17) 68 (62) 7 (18)













G 293 009 572 (1153) 994 (2176) 948 (961) ndash


G 423 lt005 23 (25)a 31 (50)ab 35 (16)b ndash

Dicots W 393 0046 91 (171) 38 (103) 44 (143) ndash


G 297 008 7 (20) 4 (08) 2 (11) ndash








G 245 011 2500 (1262) 2707 (1410) 2952 (1193) 2895 (1613)Ground cover () W 246 011 96 (04) 94 (13) 93 (07) 93 (17)





G 1076 lt001 27a 15b 12b ndash


G 581 lt005 1089 (1079)a 2235 (5178)b 2260 (4620)ab ndash


G 170 022 61 (17) 68 (36) 60 (74) ndash




G 549 lt005 14 (24)a 6 (14)b 6 (13)ab ndash


G 654 lt001 762 (1005)a 1937 (5147)b 2016 (4611)b ndash


G 835 lt001 70 (29)a 86 (26)b 88 (39)b ndash





























Treatment 2007 2011 2013












Woo

dy c

over

(

)C

hang

e in

woo

dy c

over

(

)

Fire frequency ()

0

2

4

6

8

10

12

0

0

ndash4

4

8

12

16

20

16 26 47 0 16 26 47


(a) (b)

(c) (d)





Discussion
















ndash ndash






1999 E gtL 2000 2005 B gtC








Woody cover



Season


0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000


1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)





Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)










Data collection







Woodland Grassland
















Time since fire





Results

Site vegetation



Stocking rates





Time since fire

Woody cover





0

1000

2000

3000

4000

0

20

40

60

80

100

0

20

40

60

80

0

20

40

60

Woodland

Grassland

Woodland

Grassland




Aristida latifolia


Her

bage

mas

s (k

g D

M h

andash1)

Gro

und

cove

r (

)

of h

erba

ge m

ass

o

f her

bage

mas

s

(a)

(b)

(c)

(d)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013









Grazing


Woody cover



0

5

10

15

20

25

Sto

ckin

g ra

te (

AE

km

ndash2)

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013























G 060 063 2641 (1060) 2846 (2619) 2803 (1779) 2553 (1222)Ground cover () W 216 015 79 (22) 78 (17) 68 (62) 7 (18)













G 293 009 572 (1153) 994 (2176) 948 (961) ndash


G 423 lt005 23 (25)a 31 (50)ab 35 (16)b ndash

Dicots W 393 0046 91 (171) 38 (103) 44 (143) ndash


G 297 008 7 (20) 4 (08) 2 (11) ndash








G 245 011 2500 (1262) 2707 (1410) 2952 (1193) 2895 (1613)Ground cover () W 246 011 96 (04) 94 (13) 93 (07) 93 (17)





G 1076 lt001 27a 15b 12b ndash


G 581 lt005 1089 (1079)a 2235 (5178)b 2260 (4620)ab ndash


G 170 022 61 (17) 68 (36) 60 (74) ndash




G 549 lt005 14 (24)a 6 (14)b 6 (13)ab ndash


G 654 lt001 762 (1005)a 1937 (5147)b 2016 (4611)b ndash


G 835 lt001 70 (29)a 86 (26)b 88 (39)b ndash





























Treatment 2007 2011 2013












Woo

dy c

over

(

)C

hang

e in

woo

dy c

over

(

)

Fire frequency ()

0

2

4

6

8

10

12

0

0

ndash4

4

8

12

16

20

16 26 47 0 16 26 47


(a) (b)

(c) (d)





Discussion
















ndash ndash






1999 E gtL 2000 2005 B gtC








Woody cover



Season


0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000


1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)





Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)











Time since fire





Results

Site vegetation



Stocking rates





Time since fire

Woody cover





0

1000

2000

3000

4000

0

20

40

60

80

100

0

20

40

60

80

0

20

40

60

Woodland

Grassland

Woodland

Grassland




Aristida latifolia


Her

bage

mas

s (k

g D

M h

andash1)

Gro

und

cove

r (

)

of h

erba

ge m

ass

o

f her

bage

mas

s

(a)

(b)

(c)

(d)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013









Grazing


Woody cover



0

5

10

15

20

25

Sto

ckin

g ra

te (

AE

km

ndash2)

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013























G 060 063 2641 (1060) 2846 (2619) 2803 (1779) 2553 (1222)Ground cover () W 216 015 79 (22) 78 (17) 68 (62) 7 (18)













G 293 009 572 (1153) 994 (2176) 948 (961) ndash


G 423 lt005 23 (25)a 31 (50)ab 35 (16)b ndash

Dicots W 393 0046 91 (171) 38 (103) 44 (143) ndash


G 297 008 7 (20) 4 (08) 2 (11) ndash








G 245 011 2500 (1262) 2707 (1410) 2952 (1193) 2895 (1613)Ground cover () W 246 011 96 (04) 94 (13) 93 (07) 93 (17)





G 1076 lt001 27a 15b 12b ndash


G 581 lt005 1089 (1079)a 2235 (5178)b 2260 (4620)ab ndash


G 170 022 61 (17) 68 (36) 60 (74) ndash




G 549 lt005 14 (24)a 6 (14)b 6 (13)ab ndash


G 654 lt001 762 (1005)a 1937 (5147)b 2016 (4611)b ndash


G 835 lt001 70 (29)a 86 (26)b 88 (39)b ndash





























Treatment 2007 2011 2013












Woo

dy c

over

(

)C

hang

e in

woo

dy c

over

(

)

Fire frequency ()

0

2

4

6

8

10

12

0

0

ndash4

4

8

12

16

20

16 26 47 0 16 26 47


(a) (b)

(c) (d)





Discussion
















ndash ndash






1999 E gtL 2000 2005 B gtC








Woody cover



Season


0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000


1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)





Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)








0

1000

2000

3000

4000

0

20

40

60

80

100

0

20

40

60

80

0

20

40

60

Woodland

Grassland

Woodland

Grassland




Aristida latifolia


Her

bage

mas

s (k

g D

M h

andash1)

Gro

und

cove

r (

)

of h

erba

ge m

ass

o

f her

bage

mas

s

(a)

(b)

(c)

(d)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013









Grazing


Woody cover



0

5

10

15

20

25

Sto

ckin

g ra

te (

AE

km

ndash2)

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013























G 060 063 2641 (1060) 2846 (2619) 2803 (1779) 2553 (1222)Ground cover () W 216 015 79 (22) 78 (17) 68 (62) 7 (18)













G 293 009 572 (1153) 994 (2176) 948 (961) ndash


G 423 lt005 23 (25)a 31 (50)ab 35 (16)b ndash

Dicots W 393 0046 91 (171) 38 (103) 44 (143) ndash


G 297 008 7 (20) 4 (08) 2 (11) ndash








G 245 011 2500 (1262) 2707 (1410) 2952 (1193) 2895 (1613)Ground cover () W 246 011 96 (04) 94 (13) 93 (07) 93 (17)





G 1076 lt001 27a 15b 12b ndash


G 581 lt005 1089 (1079)a 2235 (5178)b 2260 (4620)ab ndash


G 170 022 61 (17) 68 (36) 60 (74) ndash




G 549 lt005 14 (24)a 6 (14)b 6 (13)ab ndash


G 654 lt001 762 (1005)a 1937 (5147)b 2016 (4611)b ndash


G 835 lt001 70 (29)a 86 (26)b 88 (39)b ndash





























Treatment 2007 2011 2013












Woo

dy c

over

(

)C

hang

e in

woo

dy c

over

(

)

Fire frequency ()

0

2

4

6

8

10

12

0

0

ndash4

4

8

12

16

20

16 26 47 0 16 26 47


(a) (b)

(c) (d)





Discussion
















ndash ndash






1999 E gtL 2000 2005 B gtC








Woody cover



Season


0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000


1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)





Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)














Grazing


Woody cover



0

5

10

15

20

25

Sto

ckin

g ra

te (

AE

km

ndash2)

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

2011

2012

2013























G 060 063 2641 (1060) 2846 (2619) 2803 (1779) 2553 (1222)Ground cover () W 216 015 79 (22) 78 (17) 68 (62) 7 (18)













G 293 009 572 (1153) 994 (2176) 948 (961) ndash


G 423 lt005 23 (25)a 31 (50)ab 35 (16)b ndash

Dicots W 393 0046 91 (171) 38 (103) 44 (143) ndash


G 297 008 7 (20) 4 (08) 2 (11) ndash








G 245 011 2500 (1262) 2707 (1410) 2952 (1193) 2895 (1613)Ground cover () W 246 011 96 (04) 94 (13) 93 (07) 93 (17)





G 1076 lt001 27a 15b 12b ndash


G 581 lt005 1089 (1079)a 2235 (5178)b 2260 (4620)ab ndash


G 170 022 61 (17) 68 (36) 60 (74) ndash




G 549 lt005 14 (24)a 6 (14)b 6 (13)ab ndash


G 654 lt001 762 (1005)a 1937 (5147)b 2016 (4611)b ndash


G 835 lt001 70 (29)a 86 (26)b 88 (39)b ndash





























Treatment 2007 2011 2013












Woo

dy c

over

(

)C

hang

e in

woo

dy c

over

(

)

Fire frequency ()

0

2

4

6

8

10

12

0

0

ndash4

4

8

12

16

20

16 26 47 0 16 26 47


(a) (b)

(c) (d)





Discussion
















ndash ndash






1999 E gtL 2000 2005 B gtC








Woody cover



Season


0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000


1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)





Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)





















G 060 063 2641 (1060) 2846 (2619) 2803 (1779) 2553 (1222)Ground cover () W 216 015 79 (22) 78 (17) 68 (62) 7 (18)













G 293 009 572 (1153) 994 (2176) 948 (961) ndash


G 423 lt005 23 (25)a 31 (50)ab 35 (16)b ndash

Dicots W 393 0046 91 (171) 38 (103) 44 (143) ndash


G 297 008 7 (20) 4 (08) 2 (11) ndash








G 245 011 2500 (1262) 2707 (1410) 2952 (1193) 2895 (1613)Ground cover () W 246 011 96 (04) 94 (13) 93 (07) 93 (17)





G 1076 lt001 27a 15b 12b ndash


G 581 lt005 1089 (1079)a 2235 (5178)b 2260 (4620)ab ndash


G 170 022 61 (17) 68 (36) 60 (74) ndash




G 549 lt005 14 (24)a 6 (14)b 6 (13)ab ndash


G 654 lt001 762 (1005)a 1937 (5147)b 2016 (4611)b ndash


G 835 lt001 70 (29)a 86 (26)b 88 (39)b ndash





























Treatment 2007 2011 2013












Woo

dy c

over

(

)C

hang

e in

woo

dy c

over

(

)

Fire frequency ()

0

2

4

6

8

10

12

0

0

ndash4

4

8

12

16

20

16 26 47 0 16 26 47


(a) (b)

(c) (d)





Discussion
















ndash ndash






1999 E gtL 2000 2005 B gtC








Woody cover



Season


0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000


1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)





Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)















G 293 009 572 (1153) 994 (2176) 948 (961) ndash


G 423 lt005 23 (25)a 31 (50)ab 35 (16)b ndash

Dicots W 393 0046 91 (171) 38 (103) 44 (143) ndash


G 297 008 7 (20) 4 (08) 2 (11) ndash








G 245 011 2500 (1262) 2707 (1410) 2952 (1193) 2895 (1613)Ground cover () W 246 011 96 (04) 94 (13) 93 (07) 93 (17)





G 1076 lt001 27a 15b 12b ndash


G 581 lt005 1089 (1079)a 2235 (5178)b 2260 (4620)ab ndash


G 170 022 61 (17) 68 (36) 60 (74) ndash




G 549 lt005 14 (24)a 6 (14)b 6 (13)ab ndash


G 654 lt001 762 (1005)a 1937 (5147)b 2016 (4611)b ndash


G 835 lt001 70 (29)a 86 (26)b 88 (39)b ndash





























Treatment 2007 2011 2013












Woo

dy c

over

(

)C

hang

e in

woo

dy c

over

(

)

Fire frequency ()

0

2

4

6

8

10

12

0

0

ndash4

4

8

12

16

20

16 26 47 0 16 26 47


(a) (b)

(c) (d)





Discussion
















ndash ndash






1999 E gtL 2000 2005 B gtC








Woody cover



Season


0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000


1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)





Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)































Treatment 2007 2011 2013












Woo

dy c

over

(

)C

hang

e in

woo

dy c

over

(

)

Fire frequency ()

0

2

4

6

8

10

12

0

0

ndash4

4

8

12

16

20

16 26 47 0 16 26 47


(a) (b)

(c) (d)





Discussion
















ndash ndash






1999 E gtL 2000 2005 B gtC








Woody cover



Season


0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000


1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)





Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)

















Woo

dy c

over

(

)C

hang

e in

woo

dy c

over

(

)

Fire frequency ()

0

2

4

6

8

10

12

0

0

ndash4

4

8

12

16

20

16 26 47 0 16 26 47


(a) (b)

(c) (d)





Discussion
















ndash ndash






1999 E gtL 2000 2005 B gtC








Woody cover



Season


0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000


1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)





Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)










Discussion
















ndash ndash






1999 E gtL 2000 2005 B gtC








Woody cover



Season


0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000


1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)





Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)












Woody cover



Season


0

1000

2000

3000

4000H

erba

ge m

ass

(kg

DM

handash1

)H

erba

ge m

ass

(kg

DM

handash1

)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

Her

bage

mas

s of

pere

nnia

l gra

ss(k

g D

M h

andash1)

0

1000

2000

3000

4000

0

1000

2000

3000

4000

0

1000

2000

3000

4000


1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013

(a)

(b)

(c)

(d)





Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)










Frequency




0

4

8

12

16

20

0

20

40

60

80

100

0

500

1000

1500


(a)

(b)

(c)

o

f her

bage

mas

s

of h

erba

ge m

ass

Her

bage

mas

s (k

g D

M h

andash1)

1994 1995 1996 1997 1998 1999 2000 2001 2003 2005 2007 2009 2011 2013



0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)








0

200

400

600

800

1000

0

40

80

120

160

200

240

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

0

500

1000

1500

2000

2500

0

200

400

600

800

1000

1200

19941995

19961997

19981999

20002001

20032005

20072009

20112013

19941995

19961997

19981999

20002001

20032005

20072009

20112013

0

100

200

300

400

500

600

0

100

200

300

400

500

600

Her

bage

mas

s (k

g D

M h

andash1)


(a) (b)

(c) (d)

(e) (f )

(g) (h )


















































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)























































change yearndash1


008 1940s tomid-2000s


060024


017024


017024



017024







ndash






















How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)



























How often to burn



















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)

















Conclusion




Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)









Acknowledgements


References


















































































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)
























































































Season(df = 1)

Frequency(df = 3)

Year(df = 13)








How hot? How often? Getting the fire frequency and timing right for optimal management of woody...

Documents

Transcript of How hot? How often? Getting the fire frequency and timing right for optimal management of woody...