Fuel composition influences fire characteristics and understorey hardwoods in pine savanna

Fuel composition influences fire characteristics andunderstorey hardwoods in pine savannaDarin P. Ellair* and William J. Platt

Department of Biological Sciences, Louisiana State University, Baton Rouge, LA, 70803, USA

Summary

1. Fuels in the groundcover of frequently burned south-eastern pine savannas include shed leaves oftrees. Flammable needles of longleaf pine (Pinus palustris) potentially increase maximum fire tem-peratures and durations of heating, negatively affecting other trees within the groundcover. Lessflammable leaves that accumulate around the bases of understorey stems of hardwood trees such asmockernut hickories (Carya alba) in the fall potentially depress maximum fire temperatures anddurations of heating, enhancing post-fire recovery.2. We experimentally manipulated amounts of pine and hickory leaves beneath understorey hickorystems in a pine savanna, measured temperatures during prescribed fires and assessed combustion offuels and survival and regrowth of hickory stems.3. Pine needles increased fire temperatures and durations of heating relative to herbaceous fuels andincreased combustion of hickory leaves. Hickory leaves, however, neither increased nor decreasedfire characteristics relative to herbaceous fuels.4. All hickories survived fire by resprouting. When pine needles were absent, most hickories respro-uted from buds located above-ground along the stem at heights inversely related to temperatureincrease. In contrast, resprouting occurred only from underground root crowns when pine needleswere present. Such differences in locations of resprouts influenced sizes of stems at the end of thegrowing season.5. Synthesis. Groundcover fuels containing flammable leaves shed by pyrogenic species of savannatrees affect local fire characteristics and resprouting of non-pyrogenic understorey trees. Thus, localvariation in flammable fuels produced by pyrogenic species can engineer landscape dynamics ofother trees in savannas.

Key-words: Carya alba, duration of heating, ecosystem engineer, fire temperature increase, flam-mability, longleaf pine, mockernut hickory, Pinus palustris, plant population and community dynam-ics, survival and resprouting

Introduction

Fuels in fire-frequented savannas are comprised of vegetationclose to the ground surface. These fuels contain live and deadplants in the groundcover, as well as litter shed from trees inthe vicinity (Keane, Burgan & Wagtendonk 2001). As aresult, fuel type and amount vary locally depending on thevegetation present, and this variation influences fire intensity(Thaxton & Platt 2006; Hiers et al. 2009; Wenk, Wang &Walker 2011). Because overstorey trees can produce largeamounts of litter, especially in seasonal environments, theypotentially have strong local effects on the intensity of surfacefires, and hence the groundcover vegetation.Leaves shed by trees vary in flammability. For exam-

ple, needles of longleaf (Pinus palustris) and ponderosa

(P. ponderosa) pines are more flammable than needles ofsand (P. clausa) and lodgepole (P. contorta) pines (Fonda,Belanger & Burley 1998; Fonda 2001). Leaves of turkey oak(Quercus laevis) are more flammable than those of live oak(Q. virginiana) (Kane, Varner & Hiers 2008), but leaves ofboth oaks are less flammable than longleaf pine needles(Williamson & Black 1981). In general, more flammable fuelsare produced by species with leaves that are large and looselypacked when shed, that have low moisture content or thatcontain volatile oils (Dimitrakopoulos & Papaioannou 2001;Behm et al. 2004; Scarff & Westoby 2006; de Magalhães &Schwilk 2012). Such species-level variation in flammabilityshould influence local fire characteristics, especially intensity,in the vicinity of savanna trees.Differences in local fire intensity resulting from fuels with

different flammability should influence damage and survivalof understorey plants. Mutch (1970), observing that many*Correspondence author. Email: [email protected]

© 2012 The Authors. Journal of Ecology © 2012 British Ecological Society

Journal of Ecology doi: 10.1111/1365-2745.12008

plants in fire-prone ecosystems had flammable tissues,hypothesized that flammability might be beneficial to plantsin such ecosystems. Increased flammability of leaves shed bysavanna trees might open space for recruitment and survivalof their offspring, possibly by removing understorey competi-tors (Platt, Evans & Rathbun 1988; Bond & Midgley 1995).Increased flammability of leaf litter could also increasesurvival during fire by encouraging rapid consumption offuels by fast-moving fires that are less damaging to below-ground plant tissues than slower, smouldering fires (Varneret al. 2005; ‘pyrogenicity as protection’ sensu Gagnon et al.2010). In contrast to the Mutch hypothesis, however, reducedflammability might also increase survival by suppressing firebeneath trees producing such litter (Guerin 1993; Bradstock& Cohn 2002; Trauernicht et al. 2012). Reduced flammabilityof leaf litter may reduce fire intensity and allow these trees tosurvive fire intact, or allow resprouting trees to more quicklyreach a size that can withstand fire (Grady & Hoffmann 2012;Hoffmann et al. 2012).Plants indigenous to fire-frequented ecosystems have mech-

anisms for surviving fires. Overstorey trees tend to have thickor layered bark that reduces fire damage to the vascular cam-bium (Jackson, Adams & Jackson 1999). Many understoreyplants are top-killed (meaning above-ground stems are killed),but resprout from dormant meristems at ground level or onunderground storage organs (e.g. Keeley & Zedler 1978;Buchholz 1983; Ojeda, Maranon & Arroyo 1996; Peterson &Reich 2001; Drewa, Platt & Moser 2002; Higgins et al. 2007;Hoffmann et al. 2009; Werner & Franklin 2010). Local firecharacteristics, influenced by the fuels present, should influ-ence damage to dormant meristems and hence survival andregrowth of understorey plants.Effects of flammability on fire characteristics, and hence

survival of understorey plants, may differ among trees inlongleaf pine savannas. Longleaf pines (Pinus palustris Mill.),whose flammable fuels encourage fire spread, usually survivefire, while understorey hardwoods (broadleaved woody plants)are top-killed and resprout (Platt, Evans & Rathbun 1988).Mockernut hickory [Carya alba (L.) Nutt., formerly Caryatomentosa (Lam.) Nutt.] is one hardwood common in the un-derstorey of upland pine savannas in the south-eastern UnitedStates (Fig. 1a). After fires, hickories in the understoreyresprout from root crowns typically located about 5 cm belowthe soil surface. The compound leaves of the hickory areamong the largest tree leaves in pine savannas; once shed,these form a thick mat surrounding stems (Fig. 1b), possiblysuppressing fires and protecting the hickory stem and rootcrown.In this study, we explore effects of leaves shed by savanna

trees on fire characteristics and survival and resprouting ofunderstorey trees. First, we hypothesize that flammable andnon-flammable fuels produce different fire characteristics.Based on prior study and field observations of fires burningpine needles and hickory leaves, we expect characteristics offires to be augmented by pine fuels and suppressed by hick-ory fuels. We predict that increases in temperatures, durationsof heating above ambient temperature and combustion of

fuels at ground level will be modified by tree fuels added tobackground herbaceous fuels. When both pine and hickoryfuels are absent and only herbaceous fuels are present, weexpect elevation of temperatures for some time during fires(Fig. 2a). We expect lower temperatures and shorter durationsof heating based on hypothesized fire suppression effects ifhickory fuels, but no pine fuels, are added to the herbaceousfuels (Fig. 2c). We also expect further lowering of tempera-ture increases and shortening of durations of heating as theamount of hickory fuels increases in the absence of pinefuels, as more fuels with low flammability should be furthercompacted and retain more moisture (Fig. 2e). In contrast, wehypothesize that pine fuels should increase temperatures anddurations of heating above those resulting from herbaceousfuels, especially in the absence of hickory fuels (Fig. 2b). Weexpect temperatures and durations of heating between thosefor pine alone or hickory alone due to hypothesized opposingeffects of hickory and pine fuels when both are present(Fig. 2d,f). We further expect these effects of fuels on tem-perature increases and duration of heating to be reflected incombustion of tree fuels: combustion of hickory leaves should

(a)

(b)

Fig. 1. (a) Representative photograph of the study site, overstoreydominated by longleaf pine, with an understorey of bluestem grassesand scattered resprouting hardwoods; photograph taken in September,approximately 4 months since fire. (b) Photograph of understoreyhickories, showing natural litter accumulation, taken in April, 2 yearssince last fire.

© 2012 The Authors. Journal of Ecology © 2012 British Ecological Society, Journal of Ecology

2 Darin P. Ellair & William J. Platt

be increased when pine needles are present, and combustionof pine needles should be reduced when hickory leaves arepresent.Second, we hypothesize that fire characteristics produced

by different fuels should in turn influence damage and post-fire responses of understorey plants in fire-frequented ecosys-tems. We expect survival and resprouting of hickory stems tobe inversely related to temperature increases and durations ofheating. As a result, hickory stems should be less likely to bedamaged and more likely to survive fires with hickory fuelsthan fires with longleaf pine fuels.We address these hypotheses using a field experiment. We

manipulate fuel composition around hickory stems in a fre-quently burned pine savanna and measure fire characteristicsand regrowth of hickory stems. We use results of this studyto postulate how local heterogeneity in flammability of fuelsproduced by savanna trees might influence spatial heterogene-ity in savanna vegetation.

Materials and methods

We conducted our study in uplands at Girl Scouts Camp WhisperingPines (30°41′ N; 90°29′ W) in Tangipahoa Parish, Louisiana, at thewestern edge of the eastern Gulf Coastal Plain. Well-drained, fineTangi–Ruston–Smithdale Pleistocene sands mixed with and capped bydeposits of loess form a rolling topography 25–50 m a.s.l. (McDaniel1990; Platt et al. 2006). Mean annual temperature is 19 °C, and meanannual rainfall is 1626 mm (Thaxton & Platt 2006). Current overstorey

longleaf pines originated as natural regeneration within areas managedby fire for open-range grazing after widespread logging in the early1900s (Noel, Platt & Moser 1998). Fire exclusion in the 1980s resultedin increased abundance of hardwoods. Since the mid-1990s, the site hasbeen managed with biennial prescribed fires ignited during April–Mayto mimic timing of natural lightning fires (Platt et al. 2006). Longleafpines at the camp have a mean density of 293.3 ± 34.6 stems ha�1,average diameter of 28.6 ± 1.5 cm and an average basal area of18.2 ± 1.5 m2 ha�1 (Noel, Platt & Moser 1998; Platt et al. 2006; Car-michael 2012). The diverse understorey (� 30 species m�2,� 100 species 100 m�2) includes grasses, herbs, shrubs and small trees(Noel, Platt & Moser 1998; Platt et al. 2006; Gagnon et al. 2012).

We selected an area of Camp Whispering Pines with longleaf pinein the overstorey and understorey hickories in a groundcover domi-nated by warm season grasses. This area, in a burn unit scheduled fora prescribed fire in May 2010, had last been burned in May 2008.The ground was relatively level, so any effects of topography on firewere small. The burn unit was split into two separate subunits to beburned sequentially in midmorning to early afternoon on the sameday, so as to minimize likelihood of damage to campsites locatedwithin the burn unit.

We selected 30 hickory genets in January 2010, after they hadshed leaves. All had a single dominant main stem, although many(80%) also had one or more smaller stems originating from the sameroot crown. Measurements were made on the largest stem of eachgenet. All stems were resprouts since the last prescribed fire, andaverage height of the tallest stem of each genet, measured before fire,was 75 ± 4 cm. No stem had thick bark that might protect the cam-bium from fires. Bark thickness of the largest stem was 2 mm, mea-sured after the stem was top-killed by fire in 2011. Hickories werechosen far enough apart that plots would not overlap, but all 30 hick-ories were within an area about 70 m in diameter. Thus, general firecharacteristics were expected to be similar for all hickories.

Litter treatments were based on observations of litter beneath hick-ory stems. The distance of shed hickory leaves from the stems indi-cated that most hickory litter was deposited within 40 cm of thestem; this distance was used to determine plot size for litter manipula-tions. All litter present beneath stems was then removed, on average710 ± 46 g. When fuels were removed, we observed that most pine

Pine absent Pine present

Hic

kory

abs

ent

Time

Tem

pera

ture

(a)

Time

Tem

pera

ture

(b)H

icko

ry p

rese

nt

Time

Tem

pera

ture

(d)

Time

Tem

pera

ture

(c)

Mor

e hi

ckor

y

Time

Tem

pera

ture

(e)

Time

Tem

pera

ture

(f)

Fig. 2. Predicted responses of temperature–time curves for varioustreatments: (a) when both hickory and pine fuels are absent, (b) whenpine is present but hickory is absent, (c) when hickory is present butpine is absent, (d) when both pine and hickory are present in thesame amounts, (e) when a greater amount of hickory is present butpine is absent and (f) when pine is present with a greater amount ofhickory. Temperature increases (vertical arrows), and durations ofheating (horizontal arrows) are derived from logger data on changesin temperature over time at one-second intervals during fires.

0

0.1

0.2

0.3

0.4

0.5

0.6

Hickory Pine Grass Other Litter

Fuel

load

(kg

m–2

)

Fine fuelsWood

Fig 3. Fuel loads beneath hickory stems 2 years after prescribed firesat Camp Whispering Pines. Fuels derived from trees were dividedinto wood and leaves. Hickory wood consisted of dead stems killedby the last fire, while pine wood included cones, bark and branches.The ‘other’ category includes herbs, lianas and other shrub leaves,and the ‘litter’ category includes partially decomposed fragments thatwere too small to sort, predominantly of pine and hickory leaves.Bars are back-transformed least-squares means ± standard error.


Fuels influence fire and savanna shrubs 3

needles occurred beneath the layer of hickory leaves, a result of shed-ding occurring earlier in the dormant season for pines than hickories.We mimicked this layering when applying treatments.

We estimated amounts of fuels using groundcover biomass samplescollected from plots of 40 cm radius around hickory stems. Based onmean fuel load estimates of 0.1 kg m�2 (Fig. 3), an average of 50 gof hickory leaf litter was expected within a 40 cm radius of hickorystems 2 years after fire, so this amount was used as the baseline fortreatments. Fifty grams of pine needle litter was used for comparisonwith the average 50 g of hickory litter, although the average amountof pine litter beneath hickory stems at Camp Whispering Pines, basedon a mean fuel load of 0.6 kg m�2, was closer to 300 g (Fig. 3). Nowood was added to plots, as woody fuels are a minor and sporadiccomponent of the natural fuels. Wood was not present beneath allhickory stems, but where present, a single cone or fallen branch madea disproportionately large contribution to fuel load. All treatmentswere within the range of natural variation and could be found at dif-ferent locations at the study site.

Six different treatments were arranged in a 3 9 2 factorial design,with each treatment replicated five times. Circular plots with 40 cmradius (about 0.5 m2) were established around each hickory. One ofthree levels of hickory fuel was added to each plot: 0 g (0 kg m�2),50 g (0.1 kg m�2) or 100 g (0.2 kg m�2). One of two levels of pinefuel was also added to each plot: 0 g (0 kg m�2) or 50 g(0.1 kg m�2). Treatments were applied in January, 4 months beforefire, so fuels would have time to adjust to ambient conditions andbecome naturally compacted. Herbaceous plants regrew before fire, soall plots contained sufficient fine fuel to burn.

Prescribed fires were ignited on 24 May 2010, 22 days since thelast rainfall. On the day of fire, 18 plots (three replicates of each treat-ment) were burned in the first burn unit, and 12 plots (two replicatesof each treatment) were burned in the second unit. The use of twodifferent fires allowed us to observe fire characteristics under a widerrange of conditions. The first fire was ignited around 10:00 AM, whilefuels were still moist with dew, and the second fire was ignitedaround 12:30 PM when fuels were drier. Prescribed fires were ignitedusing a drip torch while walking around the burn units, first ignitingbacking and flanking fires, followed by head fire. Flame lengths wereless than 50–100 cm, and rate of spread averaged 0.6 ± 0.2 m min�1

for both fires. During burning of the first unit, air temperature was32 °C, with 66% relative humidity, and during burning of the secondunit air temperature had risen to 34 °C and relative humidity droppedto 56%. Winds were light and variable during both fires, 0–5 km h�1

from the north during burning of the first unit and 0–8 km h�1 fromthe west during burning of the second unit. Weather data wererecorded at Hammond, LA, about 18 km south of the study site, anddownloaded from the National Climatic Data Center, operated by theNational Oceanic and Atmospheric Administration (available online athttp://www7.ncdc.noaa.gov/CDO/dataproduct).

During fires, data loggers were used to record temperatures at thesoil surface beneath hickory stems at one-second intervals. Data log-gers (U12-014) and connectors (SMC-K) were obtained from OnsetComputer Corporation, Bourne, MA, and 0.81-mm thick insulatedthermocouples (XCIB-K-2-3-10) were obtained from Omega Engineer-ing, Inc., Stamford, CT. Three-metre thermocouple wires were con-nected to data loggers following general procedures in Grace, Owens& Allain (2005). Three probes were used at the base of each hickorybecause temperatures vary from point to point, and thermocouple wireshave an average 7% fail rate. The day before fire, thermocouple probeswere positioned on the soil surface beneath fuel treatments at 5 cm dis-tances and angles of 0, 120 and 240 degrees from the dominant stem.

On the morning of the fire, data loggers were attached to thermocou-ples, enclosed in plastic bags and buried well outside plots to protectthe logger from temperatures experienced during fire. We retrievedloggers about 3 h after fires. We used graphs of each temperature–timerecord to determine maximum temperature increase and duration ofheating. Temperature increase was calculated as the difference betweenthe highest temperature during fire and ambient temperature before fire(Fig. 2, vertical arrows). Duration of heating was calculated as theamount of time that temperatures remained elevated above pre-fireambient levels (Fig. 2, horizontal arrows).

After fire, we assessed survival and regrowth of hickory stems. Werecorded resprout height and resprout location for each hickory. Re-sprout height was measured as the distance above ground of the ter-minal bud of the tallest resprout. Resprout location along the stemwas measured as the distance from the ground to the base of thehighest resprout; if all resprouts originated from the underground rootcrown, this distance was zero. Measurements were recorded in Julyand September, 2 and 4 months after fire in 2010. All hickories weretop-killed by fire in the spring of 2011, so further measurements ofgrowth were not possible.

In a separate experiment, the same six treatments were repeated inopen areas of the pine savanna to measure combustion of pine andhickory fuels. Three replicates of each treatment were established inearly March in the same burn unit. Pre-fire fuels thus were 0 and50 g of pine needles and 0, 50 and 100 g of hickory leaves. Afterfire, unconsumed pine and hickory fuels were sorted from each pre-and post-fire sample, oven dried and weighed to obtain mass. Per centcombustion of pine and hickory fuel was then calculated as the ratioof unconsumed post-fire fuel mass to pre-fire fuel mass.

Analyses were conducted using SAS 9.1 (SAS Institute Inc. 2004).Temperature increases and durations of heating were analysed usingProc Mixed ANOVA. We included a random effect for plot and fixedeffects for treatments (pine and hickory) and fire unit to detect anydifferences between successive fires burned at different times of theday when fuel moistures were different. Denominator degrees of free-dom were calculated using the Kenward–Roger method. Of the 90data loggers (six treatments 9 five replicates 9 three loggers eachplot), six malfunctioned and failed to record temperatures during fireand so were excluded from analysis. Residuals were examined fornormality using Proc Univariate and the Shapiro-Wilk statistic. Tem-perature increase was natural log-transformed to correct for lack ofnormality, and no transformation was required for duration of heating.One additional outlier with respect to duration of heating wasremoved from analysis; this data point was about seven deviationsfrom the mean when residuals were analysed, and temperatures at thisdata point remained elevated for nearly 40 min and could haveresulted from burning wood falling on the thermocouple probe duringfire. ANCOVA in Proc Mixed was used to analyse relationships betweendurations of heating and temperature increases, with fuel treatmentsas covariates. Multiple regression in Proc Reg was used to detect rela-tionships between hickory regrowth (resprout location, resproutheight) and fire characteristics. We used backward variable selectionto select fire variables that best explained survival and regrowth. Onehickory that died before fire and one other outlier (five deviationsfrom the mean) were removed from analyses of regrowth. Per centcombustion was also analysed using Proc Mixed ANOVA. Combustionof pine litter was analysed with respect to hickory treatments, andcombustion of hickory litter was analysed with respect to pine treat-ments. One outlier (over four deviations from the mean) was excludedfrom analysis; fire did not burn all the way across the plot of thisoutlier.



Results

Characteristics of fires varied considerably among plots. Aver-age ambient temperature at ground level recorded just prior toignition of local fuels around thermocouples was 31 ± 0.35 °C.Maximum temperatures recorded at ground level during firesranged from 32 to 561 °C. The average maximum tempera-ture recorded across all treatments was 148 ± 11 °C, similarto values observed (147 ± 7 °C) at ground level in annuallyburned undisturbed Florida sandhill (Gibson, Hartnett & Mer-rill 1990). These maximum temperatures were lower thanthose recorded in other studies in pine savannas in whichtemperatures were recorded above ground level or with higherfuel loads (Williamson & Black 1981; Olson & Platt 1995;Drewa, Platt & Moser 2002; Kennard et al. 2005; Thaxton &Platt 2006). Durations of heating at ground level also variedamong plots, ranging from 2 to 23 min, with an average of12.9 ± 0.48 min across all treatments. These values are simi-lar to those reported (13.3 ± 1.66 min) above the ground inwiregrass patches within Carolina sandhill (Wenk, Wang &Walker 2011). These durations measured in pine savanna arealso similar to those recorded (13.0 ± 0.77 min) above theground in Ohio oak–hickory forest (Iverson et al. 2004), butare much shorter than observed (219 min) at ground level inSpanish maquis (Molina & Llinares 2001). Fire characteristicsare summarized in Table S1.Characteristics of fires were influenced by pine fuel load

but not by hickory fuel load. Mean maximum temperatureincreases at ground level were 60 °C (±1 SE ranged from 40to 92 °C) when only herbaceous fuels were present (Fig. 4a).Addition of different amounts of hickory fuels did not signifi-cantly change mean maximum temperature increases(P = 0.85; Table S2). In contrast, addition of pine fuels dou-bled temperature increases over herbaceous fuels, resulting inmean maximum temperature increases of 120 °C (±1 SE ran-ged from 79 to 182 °C; Fig. 4a); these effects were signifi-cant (P = 0.04). Interactions between pine and hickory fuelshad no significant effects on maximum temperature increases(Fig. 4a and Table S2).Durations of heating were also influenced by pine, but not

hickory fuels. Temperatures were elevated above ambient lev-els for 11.0 ± 1.3 min when herbaceous fuels alone were pres-ent (Fig. 4b). Addition of different amounts of hickory fuelsneither shortened nor lengthened durations of heating(P = 0.82). Addition of pine fuels, however, significantlylengthened (P < 0.001) durations of heating to 15.9 ± 1.3 min(Fig. 4b). Interactions between pine and hickory fuels were notsignificant (Fig. 4b and Table S3). Durations of heating werepositively related to temperature increases (P < 0.001; TableS4). Both temperatures and durations of fire were lower whenpine fuels were absent (Fig. 6a and Table S4).Successive prescribed fires in the two different burn units

produced some differences in fire characteristics at groundlevel. Maximum temperature increases were significantlylower (P = 0.03) in the earlier fire (typically about 50 °Cabove ambient temperatures) than in the later fire (typically

>100 °C above ambient temperatures). Duration of heatingwas not significantly different between the two fires. Therealso was a significant interaction between timing of fire andpine treatments (P = 0.003). During the first fire, when fuelswere moist with dew, the presence of pine litter elevated firetemperatures, but during the second fire when fuels weredrier, temperatures were generally higher regardless of thefuel present beneath hickory stems (Fig. 5). There were nosignificant interactions between hickory fuels and fire unit orcombined interactions between pine, hickory and fire unit(Table S2).All hickories survived fire, but treatments affected resprout

location along the stem. When pine fuels were present, allhickories were top-killed to the ground and resprouted frombelow-ground buds. In contrast, when pine fuels were absent,all stems were partially killed, but 78.6% resprouted fromabove-ground buds on the stem. When only herbaceous fuelswere present, 80% of hickories resprouted from above-groundbuds, on average 10.0 ± 1.0 cm above the ground. Additionof hickory fuels had no effect on resprout location. Resproutlocation also varied between the two fires. Half of the hicko-ries burned during the first fire, and 18.2% of hickoriesburned during the second fire resprouted from above-groundbuds. When pine litter was absent, all hickories burned duringthe first fire, and 40% of hickories burned during the second

0

20

40

60

80

100

120

140

160

180

200

Tem

pera

ture

incr

ease

(°C

)

0 g hickory50 g hickory100 g hickory

a

a

a

bb

b

0

2

4

6

8

10

12

14

16

18

20

0 g pine 50 g pineD

urat

ion

of h

eatin

g (m

in)

a aa

bb b

(a)

(b)

Fig. 4. (a) Effect of pine and hickory fuels on maximum increase intemperatures at ground level during prescribed fires in May 2010.Bars are back-transformed least-squares means ± standard error. (b)Effect of pine and hickory fuels on durations of heating. Bars areleast-squares means ± standard error. For both graphs, letters abovebars indicate groupings based on the main effect of pine treatment.



fire, resprouted from above-ground buds along the originalstem. Hickories that resprouted from above-ground budstended to experience both lower temperature increases andshorter durations of heating during fires (Fig. 6a).Resprout locations along the stem were related to fire char-

acteristics. As maximum temperature at ground level increased,the likelihood of resprouting from above-ground budsdecreased, and locations of above-ground resprouts shifted clo-ser to the ground (Fig. 6b). The largest increase that resulted inabove-ground resprouting was 73 °C above ambient tempera-ture (Fig. 6b). The relationship between temperature increaseand resprout location was significant (P < 0.001, R2 = 0.55;Table S5). Four months after fire, total heights of resproutingstems were still correlated with fire temperatures (P = 0.001,R2 = 0.36). Hickories that had experienced lower temperatureincreases during fire were generally taller 4 months later(Fig. 6c). Total resprout height was also somewhat correlatedwith resprout location (P = 0.02, R2 = 0.19; Table S5). Stemsthat resprouted from above-ground buds were generally taller4 months after fire (Fig. S1).Pine litter was important for the combustion of hickory lit-

ter, but hickory litter had no effects on the combustion ofpine litter. On average, 68 ± 6% of hickory litter burnedwhen no pine litter was present, but when pine litter wasadded, per cent combustion of hickory litter increased to94 ± 7% (Fig. 7a); this difference was significant(F1,9 = 9.11, P = 0.01). In contrast, hickory litter had noeffect on pine combustion (F2,5 = 1.03, P = 0.42). Onaverage, 83 ± 5% of pine needle litter burned regardless ofpresence or amount of hickory litter (Fig. 7b).

Discussion

Our study demonstrates that shed needles of longleaf pinehave large effects on fire characteristics and thus on understo-rey trees in pine savannas. Flammable needles (sensu Mutch1970; Fonda 2001) increase fire temperatures, durations ofheating and combustion of fuels, as well as top-kill understo-

rey hickories, even when fuels are moistened by dew as inthe morning fire at the study site. Most juvenile life cyclestages of longleaf pine survive groundcover fires, and so shedpine needles are likely to have greater effects on understoreyhardwoods than on understorey pines (Platt 1999). Dispersalpatterns of shed longleaf pine needles can be expected toinvolve most needles falling beneath trees. Some needles,however, have been observed to fall outside the crowns, pro-ducing distributions of needles somewhat similar to those oflongleaf pine seed dispersal (Boyer 1963; Grace, Hamrick &Platt 2004). Because flammable needles are shed over areaslarger than the crowns of trees, even low densities of needlesaway from trees should facilitate fires that maintain spacesuitable for recruitment of pines away from established trees(Platt, Evans & Rathbun 1988). Such engineering of fire

0

50

100

150

200

250

Fire unit 1 Fire unit 2

Tem

pera

ture

incr

ease

(°C

)

Pine absentPine present

a

b

b

b

Fig. 5. Interaction of fire with pine fuel treatment. Fire unit 1 burnedin the morning when fuels were moist with dew, and fire unit 2burned in the afternoon when fuels were drier. Bars are back-trans-formed least-squares means ± standard error. Letters above bars indi-cate groupings based on the interaction between fire unit and pinetreatment.

0

5

10

15

20

Res

prou

t loc

atio

n (c

m)

(b)

0

10

20

30

40

50

60

70

80

0 100 200 300

Res

prou

t hei

ght (

cm)

Temperature increase (°C)

(c)

0

5

10

15

20

Dur

atio

n of

hea

ting

(min

)

(a)

Pine absentPine presentStem survived

Fig. 6. (a) Relationship between durations of heating and temperatureincreases in plots burned during prescribed fires in May 2010. (b)Relationship between location of resprouts along stems (distanceabove the ground) and average maximum temperature increase in theplot. (c) Relationship between resprout height, measured 4 monthsafter fire, and temperature increase during fire. In all graphs, circlesrepresent plots where pine fuels were absent, and squares representplots where pine fuels were present. Filled symbols indicate plots inwhich hickory stems survived and resprouted from above-groundbuds instead of the below-ground root crown.



characteristics by overstorey longleaf pines should influenceboth stand dynamics and spatial distributions of pines andhickories in south-eastern pine savannas.Altered fire characteristics also potentially can influence

groundcover plants. Pine fuels in the vicinity of overstoreypines, often much greater than the 50 g we used in ourexperimental study and documented to reach fuel loads of600–700 g m�2 after tropical storms (Ellair 2012), shouldfacilitate frequent fires with considerably elevated tempera-tures at ground level. Pine needles also increase durationsof heating during fires, contrary to predictions of the pyro-genicity as protection hypothesis (Gagnon et al. 2010). Inthe absence of longer-burning coarse woody fuels andextreme accumulations of litter (duff), maximum tempera-tures and durations of heating should be positively corre-lated when fires are fuelled by pine needles. Increases tomaximum temperature tend to occur rapidly, and theamount of time required to cool depends on the maximumtemperatures reached (Fig. 2). In addition, the structure ofneedles and the layering of pine fuels on the ground appearto increase total time for complete combustion. Needlesrecently shed and caught in groundcover vegetation andhence located above the ground tend to combust quickly.Needles on the ground beneath vegetation (often those shedin prior years), in contrast, often combust slowly from oneend to the other after the flaming front has passed, increas-ing the duration of heating at ground level (but not above

the ground where maximum heating would have occurredduring passage of the flaming front). Longer durations ofheating at the soil surface should increase soil heating, neg-atively affecting perennial plants in the groundcover, includ-ing grasses (Gagnon et al. 2012), forbs (Wiggers 2011) andsmall trees (Thaxton & Platt 2006). We propose that theengineering of ground-level fire characteristics by pinesshould have effects on groundcover that increase with theamount of pine fuels present. Accordingly, we expect localvariation in longleaf pine fuels to result in spatial variationin composition of the groundcover. Plant species with lifecycle stages sensitive to elevated and extended heating ofthe soil surface or upper soil layers should likely belocated away from pines.Production of flammable fuels potentially could lead to

increased local biodiversity of groundcover plants. Character-istics of pine populations (sizes, density and productivity)should determine amounts and distributions of flammablecomponents of local fuels (i.e. local fuel heterogeneity; cf.Hiers et al. 2009). As a result, variation in contributions tolocal fuels could result in substantial variation in local firecharacteristics (i.e. pyrodiversity; Martin & Sapsis 1991;Faivre et al. 2011) in south-eastern savannas, especially old-growth stands with variable densities and size classes ofoverstorey pines (Platt, Evans & Rathbun 1988; Noel, Platt& Moser 1998). Such pyrodiversity should increase variationin local post-fire microenvironments (e.g. the extent to whichfuels are consumed, extent vegetation is damaged and killed,and thus, the extent to which soil is exposed post-fire). Suchvariation in fuels, and hence pyrodiversity, should in turnaffect local composition, physiognomy and dynamics of thegroundcover. Pyro-engineering by trees that produce flamma-ble fuels thus could increase biodiversity in frequentlyburned plant communities, as suggested for other ecosystems(Keeley 1990; Martin & Sapsis 1991; Faivre et al. 2011; butalso see Parr & Andersen 2006). We echo this idea, propos-ing that pyrodiversity driven by trees producing flammablefuels might enhance local heterogeneity and thus biodiversityin frequently burned, species-rich savannas such as those westudied.Hickories did not engineer fire characteristics, in contrast to

predictions that hickory fuels may suppress fires. Accordingto the pyrogenicity as protection hypothesis, flammabilityshould not benefit plants subject to large inputs of fuels fromother species. Fuels deposited by these plants contribute lessto fuel loads and are less likely to modify fire characteristics(Gagnon et al. 2010). This may be the case for hickories inpine savannas; there tend to be much more pine fuels thanhickory fuels beneath hickory stems in the vicinity of largepines. Furthermore, flammability of fuel mixtures may bedetermined by the most flammable component of the mixture(de Magalhães & Schwilk 2012). In the presence of flamma-ble fuels, fire top-kills hickory stems; hickory leaves provideno protection, either by burning rapidly or by suppressing firewhen any flammable pine fuels are present. Nonetheless,hickory stems located well away from pines are likely toresprout from above-ground buds, at least when fires occur

0

20

40

60

80

100

0 g pine 50 g pine

Com

bust

ion

of h

icko

ry le

aves

(%)

(a)

a

b

0

20

40

60

80

100

0 g hickory 50 g hickory 100 g hickory

Com

bust

ion

of p

ine

need

les

(%)

(b)

a

aa

Fig. 7. (a) Effect of pine litter on combustion of hickory leaves and(b) effect of hickory litter on combustion of pine needles. For bothgraphs, bars are least-squares means ± standard error, and letters rep-resent groupings based on main effects of litter treatments.



frequently and fuel accumulations remain low. Such hickoriesshould become taller more quickly and might eventually reacha size that they could withstand low-intensity fires.Savanna hardwoods are able to survive fires by resprout-

ing under many different fire regimes. Hardwood cover maybe reduced, but not eliminated, under short fire return inter-vals in savannas (e.g. Sankaran et al. 2005; Smit et al.2010; Werner & Franklin 2010). Even after 20 years ofannual growing season fires, some hardwoods survive byresprouting repeatedly (Waldrop, White & Jones 1992), andwe note similar results at Camp Whispering Pines afteralmost two decades of biennial prescribed fires. Increases inthe magnitude of fire characteristics can increase hardwoodtop-kill (Trollope & Tainton 1986), but higher likelihoods ofmortality tend to result primarily from fuels that produce thelargest increases in fire temperatures and burn for long peri-ods of time (Thaxton & Platt 2006). Fires early in the grow-ing season when hardwoods are investing in above-groundgrowth also are expected to limit resprouting, but notremove hardwoods from the groundcover (Glitzenstein, Platt& Streng 1995; Olson & Platt 1995; Drewa, Platt & Moser2002; Smit et al. 2010). Although fire characteristics maydamage or limit responses, once established, savanna hard-woods have a high likelihood of persisting for long periodsof time in an ‘oskar’ life cycle stage (sensu Horvitz et al.1998; an understorey plant in a suppressed juvenile state) byresprouting repeatedly. Similar patterns have been noted inother savanna trees as well (e.g. ‘gullivers’ sensu Bond &Van Wilgen 1996; ‘grubs’ sensu Peterson & Reich 2001and other trees caught in the ‘fire trap’ sensu Hoffmannet al. 2009; Grady & Hoffmann 2012). Such suppressedhardwoods may allocate more to root biomass, facilitatingstorage of photosynthates used in resprouting after shootsare killed by fire (Schutz, Bond & Cramer 2009; Tomlinsonet al. 2012).Resprouting life histories may result in persistence of hard-

woods such as hickories in south-eastern savanna landscapesdominated by ecosystem-engineering pines. The dynamics ofpopulations of resprouting hardwoods potentially could oper-ate on time-scales longer than the multiple-century life spansof reseeding pines. Establishment, then subsequent growthand survival of hardwoods should occur primarily in tran-siently open patches well away from overstorey pines, wherefire characteristics are less likely to be influenced by shedpine needles (Rebertus, Williamson & Moser 1989; Rebertus,Williamson & Platt 1993). Once established, hardwoodsshould survive as resprouting shrubs until any pines presentin the vicinity have died and conditions become favourablefor growth into the overstorey. Once hardwoods reach a sizelarge enough to escape the ‘fire trap’ (sensu Grady & Hoff-mann 2012), they may be able to reproduce and eventuallybecome large trees, as might be expected in open sandhillswhere recruitment of pines is limited (e.g. Greenberg & Si-mons 1999). Large overstorey hardwoods such as hickoriesalso may have survived on adjacent mesic slopes where fireswere of lower intensity, provided flammable fuels fromsavanna pines were not present (Platt & Schwartz 1990;

Harcombe et al. 1993). Similar patterns are expected for otherresprouting savanna trees (e.g. Higgins et al. 2007) and foresttrees that may invade savannas during fire-free intervals (e.g.Hoffmann et al. 2009).Moreover, once isolated hardwoods reach overstorey size,

they may influence their local environment within thesavanna landscape. The spreading branches and persistent (inthe fall) canopy leaves of overstorey trees such as hickorymay greatly reduce the amount of pine needles falling under-neath the canopy, protecting bases and branches of suchlarge trees from fire. The ground may become shadedenough to prevent longleaf pine establishment and reducegrowth of herbaceous fuels, enabling hardwood seedlingssuch as hickories to become established and reach a sizesufficient to survive fire. As a result, hardwoods that reachthe overstorey and suppress pine regeneration potentiallymight generate local patches of hardwoods (e.g. thickets ofHarcombe et al. 1993; oak domes of Guerin 1993), increas-ing local overstorey and understorey diversity in savannas.Similar processes might operate in other savannas, especiallyalong savanna–forest boundaries (e.g. Mitchard et al. 2009;Hoffmann et al. 2012).In summary, our study suggests that flammable fuels pro-

duced by trees could engineer aspects of local savanna land-scapes through effects on fire regimes. Local variation withinand among successive fires resulting from variation in fuelsthat influence temperatures produced and durations of heating,especially at ground level, should drive overstorey landscapedynamics involving resprouting (e.g. hickories and other hard-woods) and reseeding (e.g. longleaf pine) savanna trees.Because flammable fuels are shed into the groundcover (i.e.that component of savanna vegetation that combusts duringfires), the potential exists for such engineering of fire charac-teristics to have large effects on the plants and animals thatinhabit this layer. We propose that the vagaries of flammablefuel production and dispersal from the overstorey shouldincrease the heterogeneity of fire characteristics (pyrodiversi-ty), thereby introducing post-fire variation in environmentalconditions that potentially enhance biodiversity in fire-fre-quented savannas. Moreover, we conceptualize life historiesof non-flammable hardwood tree species like hickories asenabling them not just to persist in the presence of flammabletree species like longleaf pine, but to respond when andwhere flammable species are not present, and perhaps even toincrease locally at the expense of flammable tree species.Similar patterns might be expected in other fire-frequentedsavannas worldwide.

Acknowledgements

We thank the Girl Scouts of Eastern Louisiana for use of Camp WhisperingPines as a study site. Larry Erhlich and David Brown conducted prescribedfires. Portions of this study were supported through NSF Award 0950302(WJP, PI). Mindy Brooks, Becky Carmichael, Wynston Cormier, Daniel Depa-ula, Paul Gagnon, Leigh Griffin, Ellen Leichty, Anna Meyer and Heather Pass-more helped with the study in various ways. Rae Crandall, Thomas Dean,James Geaghan, Tracy Hmielowski, Demetra Kandalepas, Kevin Robertson,Matt Slocum, Richard Stevens and Yalma Vargas-Rodriguez provided usefulcomments on the study and manuscript.



References

Behm, A.L., Duryea, M.L., Long, A.J. & Zipperer, W.C. (2004) Flammabilityof native understory species in pine flatwood and hardwood hammock eco-systems and implications for the wildland-urban interface. International Jour-nal of Wildland Fire, 13, 355–365.

Bond, W.J. & Midgley, J.J. (1995) Kill thy neighbor – an individualistic argu-ment for the evolution of flammability. Oikos, 73, 79–85.

Bond, W.J. & Van Wilgen, B.W. (1996) Fire and Plants. Chapman & Hall,London.

Boyer, W.D. (1963) Longleaf pine seed dispersal. USDA Forest ServiceResearch Note. Southern Forest Experiment Station, New Orleans, LA.

Bradstock, R.A. & Cohn, J.S. (2002) Demographic characteristics of malleepine (Callitris verrucosa) in fire-prone mallee communities of central NewSouth Wales. Australian Journal of Botany, 50, 653–665.

Buchholz, K. (1983) Initial responses of pine and oak to wildfire in theNew Jersey Pine Barren Plains. Bulletin of the Torrey Botanical Club,110, 91–96.

Carmichael, B.J. (2012) Effects of fire and treefalls on Japanese climbing fernand native species groundcover in a restored longleaf pine savanna. PhD Dis-sertation, Louisiana State University.

Dimitrakopoulos, A.P. & Papaioannou, K.K. (2001) Flammability assessmentof Mediterranean forest fuels. Fire Technology, 37, 143–152.

Drewa, P.B., Platt, W.J. & Moser, E.B. (2002) Fire effects on resprouting ofshrubs in headwaters of southeastern longleaf pine savannas. Ecology, 83,755–767.

Ellair, D.P. (2012) Effects of hurricanes and fires on understory hardwoods inpine savannas. PhD Dissertation, Louisiana State University.

Faivre, N., Roche, P., Boer, M., McCaw, L. & Grierson, P. (2011) Character-ization of landscape pyrodiversity in Mediterranean environments: contrastsand similarities between south-western Australia and south-eastern France.Landscape Ecology, 26, 557–571.

Fonda, R.W. (2001) Burning characteristics of needles from eight pine species.Forest Science, 47, 390–396.

Fonda, R.W., Belanger, L.A. & Burley, L.L. (1998) Burning characteristics ofwestern conifer needles. Northwest Science, 72, 1–9.

Gagnon, P.R., Passmore, H.A., Platt, W.J., Myers, J.A., Paine, C.E.T. &Harms, K.E. (2010) Does pyrogenicity protect burning plants? Ecology, 91,3481–3486.

Gagnon, P.R., Harms, K.E., Platt, W.J., Passmore, H.A. & Myers, J.A. (2012)Small-scale variation in fuel loads differentially affects two co-dominantbunchgrasses in a species-rich pine savanna. PLoS One, 7, e29674, doi:29610.21371/journal.pone.0029674.

Gibson, D.J., Hartnett, D.C. & Merrill, G.L.S. (1990) Fire temperature hetero-geneity in contrasting fire prone habitats: kansas tallgrass prairie and Floridasandhill. Bulletin of the Torrey Botanical Club, 117, 349–356.

Glitzenstein, J.S., Platt, W.J. & Streng, D.R. (1995) Effects of fire regime andhabitat on tree dynamics in north Florida longleaf pine savannas. EcologicalMonographs, 65, 441–476.

Grace, S.L., Hamrick, J.L. & Platt, W.J. (2004) Estimation of seed dispersal inan old-growth population of longleaf pine (Pinus palustris) using maternityexclusion analysis. Castanea, 69, 207–215.

Grace, J., Owens, A. & Allain, L. (2005) Firelogger User’s Manual. U. S. Geo-logical Survey. National Wetlands Research Center. Fire Research Team.

Grady, J.M. & Hoffmann, W.A. (2012) Caught in a fire trap: recurring firecreates stable size equilibria in woody resprouters. Ecology, 93, 2052–2060.

Greenberg, C.H. & Simons, R.W. (1999) Age, composition, and stand struc-ture of old-growth oak sites in the Florida high pine landscape: implica-tions for ecosystem management and restoration. Natural Areas Journal,19, 30–40.

Guerin, D.N. (1993) Oak dome clonal structure and fire ecology in a Floridalongleaf pine dominated community. Bulletin of the Torrey Botanical Club,120, 107–114.

Harcombe, P.A., Glitzenstein, J.S., Knox, R.G., Orzell, S.L. & Bridges, E.L.(1993) Vegetation of the longleaf pine region of the west Gulf CoastalPlain. Proceedings of the Tall Timbers Fire Ecology Conference, 18, 83–104.

Hiers, J.K., O’Brien, J.J., Mitchell, R.J., Grego, J.M. & Loudermilk, E.L.(2009) The wildland fuel cell concept: an approach to characterize fine-scalevariation in fuels and fire in frequently burned longleaf pine forests. Interna-tional Journal of Wildland Fire, 18, 315–325.

Higgins, S.I., Bond, W.J., February, E.C., Bronn, A., Euston-Brown, D.I.W.,Enslin, B. et al. (2007) Effects of four decades of fire manipulation onwoody vegetation structure in savanna. Ecology, 88, 1119–1125.

Hoffmann, W.A., Adasme, R., Haridasan, M., Carvalho, M.T.d., Geiger, E.L.,Pereira, M.A.B., Gotsch, S.G. & Franco, A.C. (2009) Tree topkill, not mor-tality, governs the dynamics of savanna-forest boundaries under frequent firein central Brazil. Ecology, 90, 1326–1337.

Hoffmann, W.A., Geiger, E.L., Gotsch, S.G., Rossatto, D.R., Silva, L.C.R.,Lau, O.L., Haridasan, M. & Franco, A.C. (2012) Ecological thresholds at thesavanna-forest boundary: how plant traits, resources and fire govern the dis-tribution of tropical biomes. Ecology Letters, 15, 759–768.

Horvitz, C.C., Pascarella, J.B., McMann, S., Freedman, A. & Hofstetter, R.H. (1998) Functional roles of invasive non-indigenous plants in hurricane-affected subtropical hardwood forests. Ecological Applications, 8, 947–974.

Iverson, L.R., Yaussy, D.A., Rebbeck, J., Hutchinson, T.F., Long, R.P. & Pra-sad, A.M. (2004) A comparison of thermocouples and temperature paints tomonitor spatial and temporal characteristics of landscape-scale prescribedfires. International Journal of Wildland Fire, 13, 311–322.

Jackson, J.F., Adams, D.C. & Jackson, U.B. (1999) Allometry of constitutivedefense: a model and a comparative test with tree bark and fire regime. TheAmerican Naturalist, 153, 614–632.

Kane, J.M., Varner, J.M. & Hiers, J.K. (2008) The burning characteristics ofsoutheastern oaks: discriminating fire facilitators from fire impeders. ForestEcology and Management, 256, 2039–2045.

Keane, R.E., Burgan, R. & van Wagtendonk, J. (2001) Mapping wildland fuelsfor fire management across multiple scales: integrating remote sensing, GIS,and biophysical modeling. International Journal of Wildland Fire, 10, 301–319.

Keeley, J.E. (1990) Fire management for maximum biodiversity of Californiachaparral. Fire and the Environment: Ecological and Cultural Perspectives:Proceedings of an International Symposium (eds S.C. Nodvin & T.A. Wal-drop), pp. 11–14. U.S. Department of Agriculture, Forest Service, Southeast-ern Forest Experiment Station, Knoxville, TN.

Keeley, J.E. & Zedler, P.H. (1978) Reproduction of chaparral shrubs after fire:a comparison of sprouting and seeding strategies. American Midland Natu-ralist, 99, 142–161.

Kennard, D.K., Outcalt, K.W., Jones, D. & O’Brien, J.J. (2005) Comparingtechniques for estimating flame temperature of prescribed fires. Fire Ecology,1, 75–84.

de Magalhães, R.M.Q. & Schwilk, D.W. (2012) Leaf traits and litter flammabil-ity: evidence for non-additive mixture effects in a temperate forest. Journalof Ecology, 100, 1153–1163.

Martin, R.E. & Sapsis, D.B. (1991) Fires as agents of biodiversity: pyrodiversi-ty promotes biodiversity. Proceedings of the Symposium on Biodiversity ofNorthwestern California (ed. H.M. Kerner), pp. 150–157. WildlandResources Center, Division of Agriculture and Natural Resources, Universityof California, Berkeley, Santa Rosa, CA.

McDaniel, D. (1990) Soil survey of Tangipahoa Parish, Louisiana. U.S. SoilConservation Service, Washington, DC.

Mitchard, E.T.A., Saatchi, S.S., Gerard, F.F., Lewis, S.L. & Meir, P. (2009)Measuring woody encroachment along a forest–savanna boundary in centralAfrica. Earth Interactions, 13, 1–29.

Molina, M.J. & Llinares, J.V. (2001) Temperature-time curves at the soil sur-face in maquis summer fires. International Journal of Wildland Fire, 10, 45–52.

Mutch, R.W. (1970) Wildland fires and ecosystems – A hypothesis. Ecology,51, 1046–1051.

Noel, J.M., Platt, W.J. & Moser, E.B. (1998) Structural characteristics of old-and second-growth stands of longleaf pine (Pinus palustris) in the GulfCoastal region of the USA. Conservation Biology, 12, 533–548.

Ojeda, F., Maranon, T. & Arroyo, J. (1996) Postfire regeneration of a Mediter-ranean heathland in southern Spain. International Journal of Wildland Fire,6, 191–198.

Olson, M.S. & Platt, W.J. (1995) Effects of habitat and growing-season fires onresprouting of shrubs in longleaf pine savannas. Vegetatio, 119, 101–118.

Parr, C.L. & Andersen, A.N. (2006) Patch mosaic burning for biodiversity con-servation: a critique of the pyrodiversity paradigm. Conservation Biology, 20,1610–1619.

Peterson, D.W. & Reich, P.B. (2001) Prescribed fire in oak savanna: fire fre-quency effects on stand structure and dynamics. Ecological Applications, 11,914–927.

Platt, W.J. (1999) Southeastern pine savannas. The savanna, Barren, and RockOutcrop Communities of North America (eds R.C. Anderson, J.S. Fralish &J. Baskin), pp. 23–51. Cambridge University Press, Cambridge, UK.

Platt, W.J., Evans, G.W. & Rathbun, S.L. (1988) The population dynamicsof a long-lived conifer (Pinus palustris). The American Naturalist, 131,491–525.



Platt, W.J. & Schwartz, M.W. (1990) Temperate hardwood forests. Ecosystemsof Florida (eds R. Myers & J. Ewel), pp. 194–229. University of FloridaPress, Orlando, FL.

Platt, W.J., Carr, S.M., Reilly, M. & Fahr, J. (2006) Pine savanna overstoreyinfluences on ground-cover biodiversity. Applied Vegetation Science, 9, 37–50.

Rebertus, A.J., Williamson, G.B. & Moser, E.B. (1989) Longleaf pine pyroge-nicity and turkey oak mortality in Florida xeric sandhills. Ecology, 70, 60–70.

Rebertus, A.J., Williamson, G.B. & Platt, W.J. (1993) Impacts of temporal vari-ation in fire regime on savanna oaks and pines. Proceedings of the Tall Tim-bers Fire Ecology Conference, 18, 215–225.

Sankaran, M., Hanan, N.P., Scholes, R.J., Ratnam, J., Augustine, D.J., Cade,B.S. et al. (2005) Determinants of woody cover in African savannas. Nature,438, 846–849.

SAS Institute Inc. (2004) SAS 9.1.3. SAS Institute, Inc., Cary, NC.Scarff, F.R. & Westoby, M. (2006) Leaf litter flammability in some semi-aridAustralian woodlands. Functional Ecology, 20, 745–752.

Schutz, A.E.N., Bond, W.J. & Cramer, M.D. (2009) Juggling carbon: allocationpatterns of a dominant tree in a fire-prone savanna. Oecologia, 160, 235–246.

Smit, I.P.J., Asner, G.P., Govender, N., Kennedy-Bowdoin, T., Knapp, D.E. &Jacobson, J. (2010) Effects of fire on woody vegetation structure in Africansavanna. Ecological Applications, 20, 1865–1875.

Thaxton, J.M. & Platt, W.J. (2006) Small-scale fuel variation alters fire inten-sity and shrub abundance in a pine savanna. Ecology, 87, 1331–1337.

Tomlinson, K.W., Sterck, F.J., Bongers, F., da Silva, D.A., Barbosa, E.R.M.,Ward, D., Bakker, F.T., van Kaauwen, M., Prins, H.H.T., de Bie, S. & vanLangevelde, F. (2012) Biomass partitioning and root morphology of savannatrees across a water gradient. Journal of Ecology, 100, 1113–1121.

Trauernicht, C., Murphy, B.P., Portner, T.E. & Bowman, D.M.J.S. (2012) Treecover–fire interactions promote the persistence of a fire-sensitive conifer in ahighly flammable savanna. Journal of Ecology, 100, 958–968.

Trollope, W.S.W. & Tainton, N.M. (1986) Effect of fire intensity on the grassand bush components of the Eastern Cape thornveld. Journal of the Grass-land Society of Southern Africa, 3, 37–42.

Varner, J.M., Gordon, D.R., Putz, E. & Hiers, J.K. (2005) Restoring fire tolong-unburned Pinus palustris ecosystems: novel fire effects and conse-quences for long-unburned ecosystems. Restoration Ecology, 13, 536–544.

Waldrop, T.A., White, D.L. & Jones, S.M. (1992) Fire regimes for pine-grass-land communities in the southeastern United States. Forest Ecology andManagement, 47, 195–210.

Wenk, E.S., Wang, G.G. & Walker, J.L. (2011) Within-stand variation in un-derstorey vegetation affects fire behaviour in longleaf pine xeric sandhills.International Journal of Wildland Fire, 20, 866–875.

Werner, P.A. & Franklin, D.C. (2010) Resprouting and mortality of juvenileeucalypts in an Australian savanna: impacts of fire season and annual sor-ghum. Australian Journal of Botany, 58, 619–628.

Wiggers, M.S. (2011) Some like it hot: fire and legume germination in thelongleaf pine ecosystem. Master’s Thesis, Auburn University.

Williamson, G.B. & Black, E.M. (1981) High temperature of forest fires underpines as a selective advantage over oaks. Nature, 293, 643–644.

Received 18 May 2012; accepted 24 September 2012Handling Editor: Amy Austin

Supporting Information

Additional Supporting Information may be found in the online ver-sion of this article:

Fig. S1. Relationship between resprout height and resprout location.

Table S1. Summary of fire characteristics.

Table S2. Results of mixed model ANOVA on temperature increase.

Table S3. Results of mixed model ANOVA on duration of heating.

Table S4. Results of mixed model ANCOVA of duration of heating.

Table S5. Summary statistics from ANCOVA and regression analyses.



Fuel composition influences fire characteristics and understorey hardwoods in pine savanna

Documents