Content uploaded by Kimberley Simpson
Author content
All content in this area was uploaded by Kimberley Simpson on Jul 25, 2017
Content may be subject to copyright.
Content uploaded by Kimberley Simpson
Author content
All content in this area was uploaded by Kimberley Simpson on Nov 27, 2015
Content may be subject to copyright.
Determinants of flammability in savanna grass species
Kimberley J. Simpson
1
, Brad S. Ripley
2
, Pascal-Antoine Christin
1
, Claire M. Belcher
3
,
Caroline E. R. Lehmann
4
, Gavin H. Thomas
1
and Colin P. Osborne
1
*
1
Department of Animal and Plant Sciences, University of Sheffield, Sheffield S10 2TN, UK;
2
Department of Botany,
Rhodes University, PO Box 94, Grahamstown 6140, South Africa;
3
College of Life and Environmental Sciences,
University of Exeter, Exeter EX4 4PS, UK; and
4
School of GeoSciences, University of Edinburgh, Edinburgh EH9 3JN,
UK
Summary
1. Tropical grasses fuel the majority of fires on Earth. In fire-prone landscapes, enhanced flammabil-
ity may be adaptive for grasses via the maintenance of an open canopy and an increase in spa-
tiotemporal opportunities for recruitment and regeneration. In addition, by burning intensely but
briefly, high flammability may protect resprouting buds from lethal temperatures. Despite these
potential benefits of high flammability to fire-prone grasses, variation in flammability among grass
species, and how trait differences underpin this variation, remains unknown.
2. By burning leaves and plant parts, we experimentally determined how five plant traits (biomass
quantity, biomass density, biomass moisture content, leaf surface-area-to-volume ratio and leaf effec-
tive heat of combustion) combined to determine the three components of flammability (ignitability,
sustainability and combustibility) at the leaf and plant scales in 25 grass species of fire-prone South
African grasslands at a time of peak fire occurrence. The influence of evolutionary history on
flammability was assessed based on a phylogeny built here for the study species.
3. Grass species differed significantly in all components of flammability. Accounting for evolution-
ary history helped to explain patterns in leaf-scale combustibility and sustainability. The five mea-
sured plant traits predicted components of flammability, particularly leaf ignitability and plant
combustibility in which 70% and 58% of variation, respectively, could be explained by a combina-
tion of the traits. Total above-ground biomass was a key driver of combustibility and sustainability
with high biomass species burning more intensely and for longer, and producing the highest pre-
dicted fire spread rates. Moisture content was the main influence on ignitability, where species with
higher moisture contents took longer to ignite and once alight burnt at a slower rate. Biomass den-
sity, leaf surface-area-to-volume ratio and leaf effective heat of combustion were weaker predictors
of flammability components.
4. Synthesis. We demonstrate that grass flammability is predicted from easily measurable plant func-
tional traits and is influenced by evolutionary history with some components showing phylogenetic
signal. Grasses are not homogenous fuels to fire. Rather, species differ in functional traits that in
turn demonstrably influence flammability. This diversity is consistent with the idea that flammability
may be an adaptive trait for grasses of fire-prone ecosystems.
Key-words: biomass moisture content, biomass quantity, determinants of plant community diver-
sity and structure, fire regime, functional traits, phylogeny, poaceae, resprouting
Introduction
Fire is a disturbance that has shaped plant traits and floral
communities for over 420 million years (Glasspool, Edwards
& Axe 2004; Bond, Woodward & Midgley 2005) and acts as
a powerful selective filter for functional traits related to plant
persistence, recovery and recruitment (Emerson & Gillespie
2008). Fire is also multidimensional and its effects on vegeta-
tion depend on the characteristics of the local fire regime
(Keeley et al. 2011), which can vary considerably in fre-
quency, intensity, size and season (Archibald et al. 2013).
Different fire regimes can lead to the assembly of distinct
populations and communities that are functionally clustered
for diverse traits (Pausas & Bradstock 2007; Verd
u & Pausas
2007; Silva & Batalha 2010; Forrestel, Donoghue & Smith
*Correspondence author: E-mail: c.p.osborne@shef.ac.uk
©2015 The Authors. Journal of Ecology published by John Wiley &Sons Ltd on behalf of British Ecological Society.
This is an open access article under the terms of the Creative Commons Attribution License, which permits use,
distribution and reproduction in any medium, provided the original work is properly cited.
Journal of Ecology doi: 10.1111/1365-2745.12503
2014). For example, resprouting species are favoured in fre-
quent, low-intensity fire regimes, and obligate seeders that
persist via seedling recruitment are favoured in infrequent,
high-intensity fire regimes (Pausas & Bradstock 2007; Pausas
& Keeley 2014).
Plant flammability may both influence and be influenced by
fire regime (He, Lamont & Downes 2011; Pausas et al. 2012)
but species variation in flammability has received relatively
little attention (but see Scarff & Westoby 2006; Murray,
Hardstaff & Phillips 2013; Grootemaat et al. 2015). Flamma-
bility is an emergent property of a plant’s chemical and physi-
cal traits. However, the identification of these traits in several
fire-prone taxa, particularly herbaceous species, has not been
achieved. Flammability as a vegetation property consists of
several interdependent components (Anderson 1970) that can
each be quantified. Ignitability (the ease of ignition), com-
bustibility (the intensity of combustion) and sustainability (the
maintenance of burning over time) are flammability compo-
nents and can be measured at multiple scales. For example,
ignitability is often measured as ignition delay at the leaf or
plant scale, while the rate of fire spread is a measure of
ignitability that operates at the community scale (Gill &
Zylstra 2005).
Plant flammability is a key determinant of fire behaviour
(Bond & van Wilgen 1996; Beckage, Platt & Gross 2009). In
woody plants, flammability varies considerably between and
within species (e.g. Fonda 2001; Saura-Mas et al. 2010; Pau-
sas et al. 2012; Cornwell et al. 2015), and minor changes in
vegetation composition have repeatedly demonstrated signifi-
cant alterations in vegetation flammability and fire regime
(Rossiter et al. 2003; Brooks et al. 2004; Belcher et al.
2010). Flammability may act as a means by which plants
modify fire regimes to engender favourable conditions
(Schwilk 2003). For example, slow-growing, woody, obligate
seeder species, such as Pinus species, require infrequent
intense fire to complete their life cycle. High-temperature
crown fires are vital for releasing stored seeds from the
retained mature cones of these serotinous species and enhanc-
ing recruitment opportunities of seedlings via mortality of
neighbouring trees (Lamont et al. 1991; Keeley et al. 2011).
In contrast, resprouting perennial grasses, which dominate
grasslands and savannas (Uys 2000; Allan & Southgate 2002;
Overbeck & Pfadenhauer 2007), may benefit from very fre-
quent fire (Archibald et al. 2013). These shade-intolerant spe-
cies require the regular removal of standing dead biomass
(Everson, Everson & Tainton 1988) and woody growth (Bond
2008), which may be aided by high plant flammability. Sur-
face fires in grassy systems are characterized by rapid com-
bustion and spread, low fire residence times and cool burn
temperatures (Bradstock & Auld 1995; Archibald et al.
2013). Such fire characteristics are advantageous to resprout-
ing grass species, protecting basal meristems from excessive
heat through biomass that burns rapidly (Gagnon et al. 2010).
In addition, high flammability, if linked to efficient post-fire
recovery, may provide enhanced regeneration opportunities
for these species by killing neighbouring plants and reducing
post-fire competition (Bond & Midgley 1995).
Despite these predicted benefits of frequent fire to fire-
prone grasses, interspecific variation in the flammability of
such species has been little explored (Ripley et al. 2010), in
contrast to knowledge about interspecific variation in post-fire
response among grass species (Ripley et al. 2015). A histori-
cal belief persists that grasses and other herbaceous plants
vary little in their flammability, which has led to the diversity
of herbaceous fuels being reduced to one or few fuel classes
in fire behaviour modelling (e.g. Anderson 1982). Given the
considerable known variation in the flammability of woody
species (Schwilk 2003; Scarff & Westoby 2006; Pausas et al.
2012; Murray, Hardstaff & Phillips 2013), such presumptions
are unfounded. Substantial changes in grassland community
flammability resulting from invasion by non-native grasses
provide evidence to suggest considerable interspecific varia-
tion in grass flammability (Hughes, Vitousek & Tunison
1991; Rossiter et al. 2003). In addition, recent evidence
shows that grass traits relating to post-fire recovery are shaped
by fire regime (Forrestel, Donoghue & Smith 2014; Ripley
et al. 2015), suggesting that traits relating to flammability
may be responding in similar ways, resulting in intra- and
interspecific variation in flammability.
Physical and chemical traits influencing some or all compo-
nents of flammability relate to the quantity, quality, moisture
content and aeration of biomass (Bond & van Wilgen 1996;
Gill & Moore 1996). Biomass quantity is critical to com-
bustibility and fire spread rate because it directly influences
fire energy output rate (Byram 1959; Rothermel 1972). Bio-
mass moisture content determines the extent to which fuels
absorb heat energy, with high values associated with delayed
ignition and low combustion and fire spread rates (Pyne
1984; Nelson 2001). Biomass surface-area-to-volume (SA/V)
ratio influences curing and reaction rates within fires (Papio
& Trabaud 1991; Gill & Moore 1996), with high values
linked to rapid ignition, and rapid rates of combustion and
fire spread. Increasing biomass density, defined as the mass
of biomass per unit volume of fuel bed, raises fuel connectiv-
ity, therefore enhancing combustibility and fire spread rate.
This relationship applies up to a certain threshold beyond
which poor ventilation will limit drying and combustion rates
(Rothermel 1972). Intrinsic properties of plant material, such
as heat of combustion, affect combustibility and fire spread
rate through the amount of heat energy released during com-
plete combustion. Sustainability is often inversely related to
combustibility and ignitability (e.g. de Magalh~
aes & Schwilk
2012). Therefore, plant traits likely to enhance combustion
and spread rate may indirectly reduce flaming duration. In
contrast, high biomass quantity increases combustion and
spread, but is also likely to enhance sustainability, as more
fuel takes longer to burn. Plant traits important to flammabil-
ity have been identified in a number of fire-prone taxa (e.g.
Ganteaume et al. 2013; Schwilk & Caprio 2011). However,
the traits that influence grass flammability, and more generally
the flammability of herbaceous species, have not been empiri-
cally established or explored.
We examined three components of flammability, at multiple
scales, for 25 species common in fire-prone South African
©2015 The Authors. Journal of Ecology published by John Wiley &Sons Ltd on behalf of British Ecological Society., Journal of Ecology
2K. J. Simpson et al.
grasslands. Five structural and chemical plant traits, known to
influence vegetation flammability, were measured and corre-
lated with flammability trait values (see Table 1). We hypoth-
esized that (i) there is significant interspecific variation in
flammability among grass species and that (ii) the measured
plant traits can explain this variation, with each trait contribut-
ing to flammability components in different ways (see
Table 1 for specific predictions). We also expected that
flammability and plant traits covary due to the interdependent
relationships between flammability components and plant
traits. The strong phylogenetic patterns in grass distributions
across fire-frequency gradients (e.g. Visser et al. 2012; For-
restel, Donoghue & Smith 2014) led us to predict that (iii)
flammability is influenced by evolutionary history and con-
tains a phylogenetic signal.
Materials and methods
PLANT MATERIAL
Plants were collected during the natural fire season in July 2014 in
grassland and Nama-Karoo habitats near Grahamstown in the Eastern
Cape of South Africa (see Table S1 in Supporting Information for site
details). Fire return times over the 2000–2006 period were 2.3 years
for vegetation surrounding Grahamstown (Tansey et al. 2008).
Seven individuals of 25 species, representing 5 grass subfamilies,
were collected for study (see Table S2). All species were native to
the region except Cenchrus setaceus, a North African invasive species
(Milton 2004). For each species, seven randomly selected, healthy-
looking adult plants were dug up while keeping their shoot
architecture intact. Plants were stored in sealed plastic bags at room
temperature for a maximum of 48 h to minimize changes in moisture
content. A specimen of each species was deposited at the Selmar
Schonland Herbarium (Rhodes University).
STRUCTURAL AND CHEMICAL TRAITS
A section of each individual (approximately one-third of the entire
plant), with its below-ground biomass and soil removed, was used to
measure five structural and chemical plant traits. Biomass quantity,
density and moisture content were measured at the plant scale, while
effective heat of combustion (EHoC) and SA/V ratio were measured
at the leaf scale.
For measurements of leaf SA/V ratio and EHoC, leaves were
removed from a randomly selected tiller of each individual. Total leaf
area was measured on digital images using the computer program
WinDIAS (Delta-T Devices, Cambridge, U.K.) that determines leaf
area by selecting pixels of a pre-defined colour range. Leaf thickness
was measured, at the middle of the leaf and excluding the midrib, for
three leaves per tiller using digital callipers (accurate to 0.01 mm),
and an average value was calculated. Leaf SA/V ratio was calculated
from the average leaf area and leaf thickness of each species.
The heat of combustion is the energy released as heat when bio-
mass undergoes complete combustion with oxygen, which typically
relates to C:N ratio, lignin content and the presence of flammable
compounds (Philpot 1969; Bond & van Wilgen 1996). We measured
the EHoC, which is the heat of combustion of pyrolysate vapours,
and does not assume that all char is consumed. Compared to mea-
surements that involve the full thermal decomposition of biomass
(such as in bomb calorimetry), EHoC is a more realistic estimate of
the energy released from a wildfire in which combustion is incom-
plete, and most of the energy is released from burning the pyrolysate
vapours. Oven-dried leaf samples of known mass (5.0 0.4 mg)
were conditioned at room temperature and humidity before being
analysed in a microscale combustion calorimeter following the manu-
facturer’s guidelines (FAA Micro Calorimeter, Fire Testing Technol-
ogy Ltd, East Grinstead, UK). Each sample was held in nitrogen and
heated at a rate of 3 °C per second driving off the volatile gases that
were ignited and completely oxidized, and heat release was quantified
by oxygen depletion calorimetry (Tewarson 2002). Total heat release
was divided by the sample mass to provide the EHoC (kJ g
1
). Due
to the high repeatability of this trait measurement, material from three
randomly chosen individuals per species was tested in duplicate, to
give an average value per individual and per species.
For plant-scale traits, the height (maximum vertical distance from
ground level to the tallest point) and width (maximum horizontal
spread) of each clump was determined. Biomass density was mea-
sured using a novel method, which determined the vertical biomass
distribution for each individual. For this, the biomass of each clump
was divided at five or more equal intervals along its vertical height,
so that intervals were 2.5, 5, 10 or 15 cm in length depending on the
plant height, and started at ground level. Each clump was cut with
scissors at the selected intervals. The fresh and dry biomass of each
section were weighed to four decimal places, the latter after oven dry-
ing at 70 °C to a constant weight. Cumulative dry biomass was calcu-
lated at each vertical height interval from ground level. Linear models
were fitted to the logged cumulative dry biomass and vertical height
for each individual. The slope of this relationship was used as a proxy
Table 1. Matrix summarizing the predicted relationships between plant and flammability traits. Flammability traits were determined at different
scales (L, leaf; P, plant; C, community) and represent three flammability components. Symbols reflect the direction of the relationship (‘+’: posi-
tive; ‘’: negative; ‘0’: none; ‘N/A’: could not be tested). Influence is either direct or indirect (in parentheses)
Plant trait
Flammability trait
Flammability
component Scale
Biomass
quantity (g)
Biomass
density (g cm
1
)
Biomass
moisture
content (g g
1
)
Leaf
SA/V
ratio
Leaf effective
heat of
combustion (J g
1
)
Time to ignition (s) Ignitability L N/A N/A +0
Predicted rate
of fire spread (m s
1
)
Ignitability C ++ ++
Flaming time (s) Sustainability L, P +()(+)()()
Combustion rate (g s
1
) Combustibility L, P ++ ++
©2015 The Authors. Journal of Ecology published by John Wiley &Sons Ltd on behalf of British Ecological Society., Journal of Ecology
Determinants of flammability in savanna grasses 3
for biomass density, in g cm
1
, with high values indicating densely
packed biomass. For each clump, dry biomass values were combined
to give the total dry biomass, and moisture content was calculated by
dividing the difference between fresh and dry biomass by the dry bio-
mass.
FLAMMABILITY
Flammability was represented by three components: ignitability, com-
bustibility and sustainability (Anderson 1970). All components were
measured for each individual at the leaf scale via epiradiator tests. In
addition, combustibility and sustainability were determined at the
plant scale by burning partial plant canopies. Plant-scale measurement
of ignitability was beyond the scope of this experiment; however, a
community-level measure was obtained by estimating the rate of fire
spread for each individual by parameterizing Rothermel’s (1972) fire
spread model with plant trait data. Leaf- and plant-scale flammability
components were measured both on fresh and dry biomass to deter-
mine the effect of moisture content. The ‘fresh’clump was kept in a
sealed plastic bag at room temperature, and the ‘dry’clump was first
dried at 70 °C for a minimum of 48 h.
Leaf-scale ignitability, sustainability and combustibility were mea-
sured as time to ignition, flaming time and mass loss rate, respec-
tively, using a Quartz infrared 500 W epiradiator (Helios, Italquartz,
Milan, Italy) in a fume cupboard with a constant vertical windspeed
of 0.1 m s
1
. As application of leaf material directly to the epiradia-
tor’s silica disc surface always caused instantaneous combustion,
2-mm wire mesh was positioned 1 cm above the epiradiator’s surface.
The background temperature at the mesh surface (without fuel), mea-
sured by a thermocouple connected to a data-logger, ranged between
370 and 400 °C. Samples of 0.2 g (0.001 g) leaf material were cut
into 2-cm segments to standardize between samples and applied to
the centre of the mesh. The 0.2 g mass was used because preliminary
studies found that smaller masses failed to ignite, while larger fuel
masses increased the risk that other fuel properties, particularly fuel
height, influenced flammability values. Smaller samples were used for
Aristida congesta subsp. barbicollis due to the low leaf mass of this
species. Each test was filmed at 25 frames s
1
, and (i) time to igni-
tion (TTI; the time between sample application to the epiradiator and
first flaming) and (ii) flaming time (FT; the time from ignition to
flame extinction) were subsequently determined. As samples were
completely combusted by applying them to the epiradiator, an average
leaf combustion rate was obtained by dividing the mass of samples
by FT. Species average values for TTI and FT were obtained for
fresh and dry material. The influence of leaf moisture content on
these flammability traits was determined as the difference in values
between fresh and dry samples of each individual and averaged per
species.
As canopy architecture influences grass flammability (Martin
2010), a method that measures plant-scale flammability traits was uti-
lized. Fresh and dry plant material from each individual were clamped
on a stand on a four-point balance (Mark 205A; Bel Engineering,
Monza, Italy) and burnt in a fume cupboard with a constant
0.1 m s
1
vertical wind speed (see Figure S1 for diagram of the set-
up used). Samples were ignited by directing a Bunsen burner flame to
the side of the base of the clump at a 45°angle and a 5 cm distance
for a maximum of 3 s (less if ignition happened earlier). This resulted
in successful ignition in all individuals. Mass loss was logged at 0.2-s
intervals and the sigmoidal relationship produced was fitted with a
Boltzmann equation. Data were excluded if fitting the relationship
was not possible due to noise around the curve (n=40/350), which
occurred if large pieces of plant material fell off the balance during a
burn. The width parameter used to fit the Boltzmann curve reflects
the time period in which mass was drastically reduced and was used
as a plant-scale measurement of sustainability (flaming time). Three
seconds of data either side of the inflection point were selected and a
linear regression fitted. The slope of this regression represents the
maximum combustion rate in g s
1
. As preliminary results found this
combustibility trait to be strongly driven by the biomass of the sam-
ple, interspecific comparisons were standardized for mass. Therefore,
maximum combustion rate was plotted against mass change for each
species, and linear models were fitted to the fresh, dry and combined
data sets. As there was no change in mass common to all 25 species,
the y-intercept extracted from the model fitted to the combined data
set was used to characterize the intrinsic combustibility of each spe-
cies. The combined data set was used as the slopes of the models fit-
ted to the fresh and dry data did not differ significantly for any
species, and model fit was improved by combining the data sets. Any
unpaired samples were excluded to ensure a balanced data set of fresh
and dry samples. The y-intercept differed significantly between fresh
and dry models for three species (Panicum sp., Hyparrhenia hirta
and Merxmuellera stricta) and in these cases, the y-intercept was
extracted from linear models fitted to the fresh data set.
Forward fire spread rate values, the community-scale measure of
ignitability, were predicted for each individual using Rothermel’s
(1972) surface fire spread model as implemented using the ros() func-
tion in the Rothermel package (Vacchiano & Ascoli 2014) in R (R
Core Team 2013). Fire behaviour was simulated for each individual
by parameterizing the model with data for the following traits: leaf
SA/V ratio, leaf EHoC, biomass moisture content, plant height and
fuel load (biomass quantity divided by the estimated cover area). See
Table S3 for a details of the procedure followed and model assump-
tions.
PHYLOGENETIC ANALYSIS
We constructed a phylogeny that was initially based on a previously
generated data set for grasses composed of the plastid markers
trnKmatK,ndhF and rbcL (Grass Phylogeny Working Group II 2012)
and augmented here. For ten species not represented in this previous
data set, a fragment of trnKmatK was PCR-amplified from genomic
DNA, following protocols and primers described previously (Grass
Phylogeny Working Group II 2012). The newly generated sequences
have been submitted to NCBI database (Benson et al. 2012) under
the accession numbers KP860326 to KP860336. The new markers
were manually aligned to the data set, which consisted of 606 taxa
and 5649 aligned bp. This initial data set was downsized to 70 spe-
cies, including all the taxa studied here and representatives of all
grass lineages. A time-calibrated phylogenetic tree was obtained
through Bayesian inference as implemented in BEAST (Bayesian evo-
lutionary analysis by sampling trees; Drummond & Rambaut 2007).
A general time-reversible substitution model with a gamma-shape
parameter and a proportion of invariants (GTR+G+I) were used. The
log-normal relaxed clock was selected. The tree prior was modelled
by a Yule process. The monophyly of the BEP-PACMAD clade was
enforced, leaving Puelia olyriformis as the outgroup. The calibration
prior for the age of the BEP-PACMAD crown was set to a normal
distribution, with a mean of 51.2 and a standard deviation of 0.001
(mean based on Christin et al. 2014). Two independent runs were
conducted for 10 000 000 generations, sampling a tree every 1000
generations. The convergence of the runs and the appropriateness of
the burn-in period, set to 2 000 000 generations, were verified using
©2015 The Authors. Journal of Ecology published by John Wiley &Sons Ltd on behalf of British Ecological Society., Journal of Ecology
4K. J. Simpson et al.
Tracer (Rambaut A, Drummond AJ (2007) Tracer v1.4, available at
http://beast.bio.ed.ac.uk/Tracer). Median ages were mapped on the
maximum-credibility tree. The relationships among the species studied
here were extracted from this tree and used for comparative analyses.
DATA ANALYSIS
Statistical analyses were carried out in the R environment (R Core
Team 2013). Data were log-transformed to improve normality and to
meet model assumptions where necessary.
Analysis of variance (ANOVA) was used to determine whether plant
and flammability traits differed significantly between species. The influ-
ence of species, and state (‘fresh’or ‘dry’), on leaf-scale flammability
was determined by two-way ANOVA. As biomass quantity values for the
plant-scale burns are not representative of the species (i.e. for each spe-
cies, clumps were subsampled and a range of masses were burnt), a spe-
cies effect on the relationship between maximum combustion rate and
biomass quantity was tested using the R package MCMCglmm (Had-
field 2010). This approach implements Markov chain Monte Carlo rou-
tines for fitting generalized linear mixed models, while accounting for
non-independence and correlated random effects arising from phyloge-
netic relationships (Hadfield 2010). We fitted flammability (maximum
combustion rate) and biomass quantity as a bivariate normal response,
and species as a random effect. Models were run for 500 000 iterations
with a burn-in of 1000 iterations, a thinning interval of 500 and weakly-
informative priors (V=diag(2), nu =0.002). The 95% highest poste-
rior densities (HPD) of within-species and across-species slopes and the
difference between slopes were estimated while accounting for phy-
logeny and used to assess whether slopes differed among species.
To test the hypotheses put forward in Table 1 and to establish the
strength and direction of plant trait contributions to flammability com-
ponents, a MCMC multi-response generalized linear mixed model
approach was used again. Traits were separated into leaf and plant
scale to ensure appropriate comparisons, using the same prior and
specifications as before. The fit of the models to data was established
by fitting linear models between the observed flammability trait val-
ues and those predicted by the models. The contribution of plant traits
to fire spread rate was tested to determine whether strong relation-
ships occurred across species when accounting for phylogeny, while
acknowledging that some circularity is involved because spread rate
was predicted based on the values of these traits.
To explore the pattern of covariance among plant and flammability
traits, principal component analyses were performed using the prin-
comp function (R core team 2013). Linear regressions were also used
to establish the relationships among plant and flammability traits, with
the latter being split into leaf-scale and plant-scale traits for analyses
to ensure an appropriate comparison. The relationships between
flammability traits measured at different scales were also established
using linear regressions.
The influence of evolutionary history was established for each
plant and flammability trait by testing for the presence of a phyloge-
netic signal. This was done using the pgls function in the caper pack-
age (Orme et al. 2012) which estimated Pagel’sk.
Results
FLAMMABILITY VARIATION AMONG SPECIES
All flammability components varied considerably across spe-
cies (Fig. 1; Table S4). At the leaf-scale, significant inter-
specific variation was found in ignitability (F
24,144
=5.02,
P<0.001), sustainability (F
24,144
=3.02, P<0.001) and
combustibility (F
24,144
=2.97, P<0.001). Ignition delays
ranged from 1.0 s (H. hirta) to 4.0 s (C. setaceus) with a
mean across species of 1.7 s. The mean flaming duration
across species was 6.3 s and ranged from 4.3 s (A. congesta
subsp. barbicollis) to 7.6 s (Eragrostis plana). Connected to
flaming duration was average combustion rate, with E. plana
burning at the slowest rate (27 mg s
1
) and A. congesta
subsp. barbicollis at the fastest (49 mg s
1
).
At the plant scale, intrinsic combustibility (for a given bio-
mass) differed by <2.5-fold across species, ranging from
0.064 g s
1
(Eustachys paspaloides) to 0.163 g s
1
(The-
meda triandra). When investigating the relationship between
combustion rate and biomass, the bivariate mixed effects
model revealed that within-species slopes (pooled
mean =0.594, HPD: 0.507 to 0.707) and across-species
slopes (mean =0.797, HPD: 0.067 to 1.385) did not differ
significantly (mean slope difference (Db) =0.212, HPD:
0.521 to 0.683) when accounting for phylogeny (Fig. 2).
This common relationship was extrapolated while taking into
account intrinsic combustibility differences, allowing combus-
tion values to be predicted for the species mean total biomass.
These predicted values of whole-plant combustion rates varied
>20-fold among species, ranging from 0.06 g s
1
(A. con-
gesta subsp. barbicollis) to 1.28 g s
1
(M. disticha; Fig. 2).
Fuel models based on the traits of C. setaceus predicted no
fire spread, because biomass moisture content values
exceeded the moisture of extinction, defined as the fuel mois-
ture content above which a steady rate of fire spread is not
possible. Of the remaining species that spread fire, the esti-
mated rate of spread differed substantially (25-fold; Table S4)
and varied significantly between species (ANOVA:
F
24,150
=42.42, P<0.001).
Substantial interspecific variation was also found in the five
traits measured as explanatory traits for flammability (Fig. 1;
see Table S5). Biomass moisture content values of the non-
native C. setaceus were substantially higher than the other
species. However, species still differed significantly for this
trait when C. setaceus was excluded (ANOVA:F
23,144
=14.39,
P<0.001). The measurement of biomass density (i.e. vertical
biomass distribution) produced consistent values within spe-
cies (Fig. S2; species average CV =28%), but considerable
differences among species with slope values ranging from
0.155 (Eragrostis lehmanniana) to 0.831 (M. stricta).
Collection site did not influence flammability traits. Of the
plant traits, vertical biomass distribution (P=0.008) and leaf
EHoC (P=0.046) were the only ones affected by collection
site (see Table S7).
TRAIT CONTRIBUTIONS TO FLAMMABILITY
Measured plant traits significantly predicted the components
of flammability, particularly ignitability and plant-scale com-
bustibility, in which 70% and 58% of variation could be
explained by the plant traits, respectively (Tables 2 and 3).
Variation in sustainability could be explained to a lesser
extent by plant traits at the leaf (47%) and plant scale (37%),
©2015 The Authors. Journal of Ecology published by John Wiley &Sons Ltd on behalf of British Ecological Society., Journal of Ecology
Determinants of flammability in savanna grasses 5
as well as variation in leaf-scale combustibility (39%). The
direction of relationships between plant and flammability
traits is consistent with those predicted in Table 1, but there
are exceptions. Both biomass density and leaf SA/V ratio
were expected to correlate positively with predicted spread
rate, but instead correlated negatively (Table 3).
Moisture content was key in determining leaf-scale flamma-
bility components (Table 2; Table S6). Ignitability was partic-
ularly influenced by moisture content, with fresh leaf material
taking 42% longer to ignite on average than dry leaf material
across species, with a maximum increase of 288% seen for C.
setaceus (1.0 s dry vs. 4.0 s fresh). Once alight, fresh leaf
material also burned on average for 7% longer at a 3% lower
combustion rate compared to dry leaf material across species.
Leaf SA/V ratio significantly influenced sustainability, with
high values associated with low flaming duration. The EHoC
of leaf material alone contributed little to overall leaf-scale
flammability when compared to moisture or SA/V ratio
(Table 2).
At the plant scale, biomass quantity was by far the stron-
gest driver of sustainability and combustibility (Table 3).
Plants with greater biomass burnt at a faster rate and for
longer. Biomass density and moisture content significantly
Fig. 2. Relationships between biomass quantity and maximum com-
bustion rate across 25 grass species. The mean slopes of within-
species relationships (grey lines) and across-species relationships
(black dotted line) for maximum combustion rate with biomass
burned do not differ significantly when phylogeny is accounted for.
Data points are shown as grey circles. Estimates of whole-plant com-
bustion rates (black diamonds) showed substantial variation (>20-
fold). These values were calculated by extrapolating the common
across-species relationship (black dashed line) to species mean total
biomass values while taking into account the intrinsic combustibility
differences among species.
Fig. 1. The evolutionary relationships between species and average values of explanatory plant traits (solid circles) and flammability traits (open
circles). Trait values are indicated by the size of the circles. A nonzero phylogenetic signal was found for leaf SA/V ratio (Pagel’sk=1; P=1
for k=1; P<0.001 for k=0), leaf flaming time (Pagel’sk=0.45; P=1.0 for k=1; P<0.001 for k=0) and leaf combustion rate (Pagel’s
k=0.99; P=0.93 for k=1; P=0.037 for k=0).
©2015 The Authors. Journal of Ecology published by John Wiley &Sons Ltd on behalf of British Ecological Society., Journal of Ecology
6K. J. Simpson et al.
contributed to plant-scale combustibility, such that plants with
high density and low moisture content combusted most
rapidly (Table 3). The EHoC of leaf material significantly
contributed to sustainability with high values associated with
short flaming times (Table 3). Leaf SA/V ratio did not signifi-
cantly contribute to plant-scale combustibility or sustainabil-
ity.
Biomass load, moisture content, density and leaf SA/V
ratio all contributed highly to predicted fire spread rate when
taking phylogeny into account (Table 3). Fuel load con-
tributed directly to reaction intensity and indirectly to the
propagating flux ratio, via bulk density. Biomass moisture
content contributed to spread rate by increasing the heat
required for ignition and damping the reaction intensity (see
Fig. S2). Leaf SA/V ratio influenced reaction intensity and
the proportion of this reaching adjacent fuel (propagating flux
ratio), as well as the proportion of fuel raised to ignition tem-
perature (effective heating number; Fig. S2). Leaf EHoC con-
tributed to the reaction intensity but played a small part in
determining the overall predicted rate of spread (Table 3;
Fig. S2).
TRAIT COVARIANCE
Principal components analysis (PCA) and linear regressions
were used to explore patterns of covariance among the plant
and flammability trait variables, with the latter being split into
leaf-scale and plant-scale traits (Fig. 3). For the plant traits,
the first two principal components accounted for 67.6% of the
total variance. The first axis related to the chemical properties
of biomass and how it is arranged spatially (leaf EHoC, bio-
mass moisture content and density had the highest axis load-
ings). Leaf SA/V ratio loaded most heavily on the second
axis, followed by biomass moisture content and density. Only
biomass quantity did not fall as clearly on the first two princi-
pal components, which we believe is due to the high variation
within the data (CV =89.0%). For the leaf-scale flammability
traits, the first two principal components accounted for 95.1%
Table 2. The contribution of plant traits to leaf-scale flammability components as determined by MCMC phylogenetic generalized linear mixed
models. Values represent posterior mean estimates of the slopes, the upper and lower 95% confidence intervals and Pvalues (those in bold are
significant at P=0.05). In combination, species mean trait values of leaf moisture content, SA/V ratio and effective heat of combustion (EHoC)
significantly predicted ignitability (F
1,166
=398.3, P<0.001, R
2
=0.70), sustainability (F
1,166
=147.5, P<0.001, R
2
=0.47) and combustibility
(F
1,166
=105.4 P<0.001, R
2
=0.39)
Leaf moisture content* Leaf SA/V ratio log Leaf EHoC
Ignitability (time to ignition) Estimate 0.691 0.174e-3 0.135e-4
(95% CI) (0.620 to 0.760) (0.420e-3 to 0.872 e-5) (0.527e-4 to 0.290e-4)
Pvalue <0.001 0.17 0.49
Sustainability (flaming time) Estimate 0.492 0.876e-3 0.159e-4
(95% CI) (0.421 to 0.567) (0.142e-2 to -0.359 e-4) (0.626e-4 to 0.113e-3)
Pvalue <0.001 0.002 0.741
Combustibility (combustion rate) Estimate 0.303e-2 0.522e-5 0.227e-6
(95% CI) (0.406e-2 to 0.170e-2) (0.547e-5 to 0.164e-4) (0.254e-5 to 0.193e-5)
Pvalue <0.001 0.36 0.86
*Parameter characterized as: the species mean difference in ignition delay (for ignitability) or flaming duration (for sustainability and combustibil-
ity) between fresh and dry leaf material for each individual.
Table 3. The contribution of plant traits to plant-scale flammability components as determined by MCMC phylogenetic generalized linear mixed
models. Values represent posterior mean estimates of the slopes, the upper and lower 95% confidence intervals and Pvalues (those in bold are
significant at P=0.05). Values represent posterior mean estimates of the slopes, the upper and lower 95% confidence intervals and Pvalues
(those in bold are significant at P=0.05). In combination, the five plant traits significantly predicted sustainability (F
1,151
=90.07, P<0.001,
R
2
=0.37), combustibility (F
1,151
=210.8, P<0.001, R
2
=0.58) and ignitability (F
1,173
=184.2, P<0.001, R
2
=0.51)
log Biomass
quantity
log Biomass
density
log Biomass
moisture content Leaf SA/V ratio log Leaf EHoC*
Sustainability
(flaming time)
Estimate 0.434 0.614 1.036 0.050 0.012
(95% CI) (0.350 to 0.517) (2.162 to 0.889) (0.688 to 2.753) (0.162 to 0.055) (0.023 to 0.001)
Pvalue <0.001 0.443 0.252 0.363 0.060
Combustibility
(maximum
combustion rate)
Estimate 0.035 0.149 0.108 0.105e-2 0.580e-4
(95% CI) (0.028 to 0.041) (0.021 to 0.277) (0.250 to 0.027) (0.858e-2 to 0.012) (0.101e-2 to 0.103e-2)
Pvalue <0.001 0.024 0.116 0.910 0.826
Ignitability
(predicted
spread rate)
Estimate 2.002 0.061 0.034 0.128e-2 0.121e-3
(95% CI) (0.951 to 3.015) (0.094 to 0.033) (0.044 to 0.025) (0.789e3 to 0.169e-2) (0.993e-4 to 0.360e-3)
Pvalue <0.001 <0.001 <0.001 <0.001 0.309
*Species mean values.
©2015 The Authors. Journal of Ecology published by John Wiley &Sons Ltd on behalf of British Ecological Society., Journal of Ecology
Determinants of flammability in savanna grasses 7
of the total variance. Leaf flaming time and combustion rate
were negatively correlated (P<0.001), and fell in opposing
directions on the first PCA axis (Fig. 3), which reflects how
combustion rate was derived from flaming time. Time to igni-
tion was unrelated to flaming time and combustion rate and
was orthogonal to both in the PCA (Fig. 3). For plant-scale
flammability traits, 71.8% of total variance is accounted for
by the first two principal components. Traits did not separate
on the first axis, but did on the second axis which related to
burning intensity. High rates of plant combustion were associ-
ated with rapid predicted fire spread rates (P<0.001) and
marginally with longer flaming times (P=0.071; Fig. 3).
The relationships between flammability traits measured at
different scales were variable, with a significantly positive
correlation found for ignitability (leaf time to ignition vs.
predicted rate of spread; P=0.025), but no significant corre-
lation for combustibility (leaf-scale combustion rate vs.
plant-scale combustion rate; P=0.29).
INFLUENCE OF EVOLUTIONARY HISTORY ON
FLAMMABILITY
Support for a phylogenetic signal was found for leaf-scale
combustibility (Pagel’sk=0.99; P=0.93 for likelihood
ratio test against k=1; P=0.037 against k=0) and sustain-
ability (Pagel’sk=0.45; P=0.67 against k=1; P=0.011
against k=0), but not for the other flammability traits. Of
the plant traits, there was a strong phylogenetic signal for leaf
SA/V ratio (Pagel’sk=1.00; P=1.00 against k=1;
P<0.001 against k=0), with closely related species tending
to have similar values of leaf SA/V ratio. No phylogenetic
signal was found in the other plant traits.
Discussion
This large comparative study of grass flammability provides
strong support for the hypothesis that grass species vary
significantly in multiple components of flammability. This find-
ing suggests that static classifications of grassy and herbaceous
vegetation as homogenous fuels mask considerable interspecific
and community variation in flammability. Consequently, fire
behaviour predictions based on such fuel models may lose
accuracy when community composition is not accounted for.
A substantial proportion of variation in ignitability and
combustibility (70% and 58%, respectively) can be explained
by a combination of the five plant traits measured here. For
sustainability, a smaller proportion of variation was accounted
for (37%), perhaps because this component is not only driven
by plant traits, but is also directly influenced by combustibil-
ity. Additionally, some variation in sustainability could be
accounted for by traits relating to leaf chemistry, such as
nitrogen, phosphorus and tannin concentrations (Grootemaat
et al. 2015) that were not measured in this study. Biomass
quantity was the key trait influencing plant-scale flammability
components and also determined the influence of an individ-
ual plant on local fire characteristics. The importance of bio-
mass quantity for combustibility, sustainability and fire spread
rates in the field is illustrated by the elevated flammability of
landscapes caused by the raised fuel load production of non-
native grasses (Hughes, Vitousek & Tunison 1991; D’Antonio
& Vitousek 1992; Rossiter et al. 2003). While making a rela-
tively small contribution to flammability components once
alight, biomass moisture content was key to ignitability, with
higher moisture contents requiring more energy to dry and
heat biomass to the point of ignition (Trollope 1978; Gill &
Moore 1996; Alessio et al. 2008; Plucinski & Anderson
2008). By influencing ignitability, and therefore the likelihood
of fire occurring in the first place, moisture content exerts a
strong influence on vegetation flammability. Our finding of
high interspecific variation in EHoC (effective heat of com-
bustion) also conflicts with the notion that grass energy con-
tent is an almost constant value (Trollope 1984). However,
EHoC contributed little to leaf-scale flammability components,
supporting the idea that this intrinsic property is less
Fig. 3. Principal components analysis biplots of explanatory plant traits (a) and flammability traits at the leaf scale (b) and plant scale (c). The
tables within each plot indicate the slope and significance of linear regressions between each pair of variables. Data for all traits were log-
transformed to improve normality except leaf SA/V ratio. EHoC is the leaf effective heat of combustion. P<0.1; *,P<0.05; ***,P<0.001.
©2015 The Authors. Journal of Ecology published by John Wiley &Sons Ltd on behalf of British Ecological Society., Journal of Ecology
8K. J. Simpson et al.
important in determining flammability than fuel mass, struc-
ture and moisture content (Bond & van Wilgen 1996).
Despite this small importance overall, the EHoC marginally
contributed to plant-scale flaming time.
The inconsistent relationships between components of
flammability, and within flammability components measured
at different scales, suggest that descriptions of flammability
should incorporate all suitable components and should be
taken at an appropriate scale. The mixed covariance between
flammability components found here suggests that one cannot
always be used as a proxy for the others. Therefore, studies
that consider one or even two components of flammability
may mask the complexity of vegetation flammability (Ander-
son 1970). Similar to the findings of Martin (2010), we find
support for the importance of incorporating plant architecture
into measurements of grass flammability. Inconsistencies
between combustibility at the leaf- and plant-scale highlight
that other factors (such as biomass quantity and density) are
key determinants of combustibility at the plant scale. Bench-
scale measurements of flammability have been criticized as
not being representative of flammability in the field (Fernan-
des & Cruz 2012), and our findings emphasize the need for
caution when extrapolating flammability traits between differ-
ent scales. In comparison with leaf-scale studies, the flamma-
bility component values obtained here are more representative
of flammability in the field because they are measured at the
plant scale and on field-state plants that are at the phenologi-
cal stage most appropriate to fire occurrence.
The phylogenetic signal found in some flammability com-
ponents (leaf-scale combustibility and sustainability) suggests
that evolutionary history may partially explain patterns of
grass flammability and the strong sorting of grass lineages
across fire-frequency gradients (Uys, Bond & Everson 2004;
Visser et al. 2012; Forrestel, Donoghue & Smith 2014). How-
ever, conclusions on phylogenetic signal derived from a small
phylogeny must remain cautious due to low statistical power
(Boettiger, Coop & Ralph 2012).
Through their flammability, plants may modify the fire
regime they experience in order to increase their fitness in
fire-prone environments (Schwilk 2003). Resprouting grasses
are likely to benefit from frequent fires that remove standing
biomass and maintain an open canopy, because they are typi-
cally intolerant of shading (Everson, Everson & Tainton
1988; Bond 2008). The grasses studied here showed high
ignitability, combustibility and predicted fire spread rates,
when compared to woody vegetation fuels (e.g. Pausas et al.
2012; Ganteaume et al. 2013). Furthermore, grasses are able
to regrow quickly after fire. This combination of high
flammability and rapid regrowth drives a fire regime charac-
terized by high fire frequency (Grigulis et al. 2005). Plant-
scale combustion rate was marginally positively related to
flaming time, with high biomass plants burning at a faster rate
and for longer. This finding is in contrast with other studies
(e.g. de Magalh~
aes & Schwilk 2012) that found a negative
relationship between the two. It also does conflicts with the
idea of high flammability providing resprouting plants protec-
tion against lethal temperatures (Gagnon et al. 2010), as for
grasses that have higher fuel loads, rapid combustion is not
associated with lowered burning durations and a subsequent
reduction in heat transfer to the soil and below-ground plant
parts. The interspecific variation in flammability components
observed across a set of species that commonly coexist in the
field further suggests a role for interspecific competition in
promoting flammability as an adaptive trait. Potentially,
enhanced plant flammability can increase the mortality of
neighbouring, less fire-tolerant individuals and thereby reduce
post-fire competition (Bond & Midgley 1995). Furthermore,
some evidence provides intriguing support for a link between
high flammability and ecological success in fire-prone grass-
land species (Ripley et al. 2015). The influence of flammabil-
ity at the species level on grassland community-level
flammability has not been determined. However, findings
from other vegetation fuel types show that flammability tends
to be driven by the most flammable species of a community,
such that fuel loads are non-additive (van Altena et al. 2012;
de Magalh~
aes & Schwilk 2012). The knowledge gained in
this study can be used in further work to determine whether
high flammability is an adaptation to life in frequently burnt
environments for grasses and has thus been a fundamental
trait in grass evolution. In addition, the knowledge of inter-
specific variation in grass flammability obtained here can lead
to a better understanding of wildfire behaviour, particularly in
grassland ecosystems. This could potentially contribute to an
improvement of global carbon modelling and lead to new
insights about ecosystem feedback to fire.
Acknowledgements
Research support was provided by a Natural Environment Research Council
studentship to K.J.S., Royal Society University Research Fellowship
URF120119 to P.A.C. and URF120016 to G.H.T. and a European Research
Council Starter Grant ERC-2013-StG-335891-ECOFLAM to C.M.B. Author
contributions: K.J.S., G.H.T., B.S.R., C.M.B., C.E.R.L. and C.P.O. designed
the study. K.J.S., B.S.R. and P.A.C. generated the data. K.J.S., P.A. C., B.S.R.,
G.H.T. and C.P.O. analysed the data. K.J.S. wrote the manuscript with the help
of all the authors. We thank Tony Palmer, Claire Adams and Nosipho Plaatjie
for their support in the laboratory and field, Albert Phillimore for assistance
with the MCMCglmm analyses and James Simpson for his help with graphics.
We also thank Hans Cornelissen and two anonymous referees for their con-
structive comments on the manuscript.
Data accessibility
Trait data: Species average values uploaded as online supporting information;
raw data available in DRYAD entry doi: 10.5061/dryad.2c506.
Sequence data: GenBank accession numbers available as online supporting
information.
Phylogeny: Nexus file available in DRYAD entry doi: 10.5061/dryad.2c506.
MCMCglmm R Script: Available in DRYAD entry doi: 10.5061/
dryad.2c506.
References
Alessio, G.A., Penuelas, J., Llusia, J., Ogaya, R., Estiarte, M. & De Lillis, M.
(2008) Influence of water and terpenes on flammability in some dominant
Mediterranean species. International Journal of Wildland Fire,17, 274–286.
Allan, G.E. & Southgate, R.I. (2002) Fire regimes in the spinifex landscapes of
Australia. Flammable Australia: the Fire Regimes and Biodiversity of a Con-
tinent (eds R.A. Bradstock, J.E. Williams & A.M. Gill), pp. 145–176.
Cambridge University Press, Cambridge, UK.
©2015 The Authors. Journal of Ecology published by John Wiley &Sons Ltd on behalf of British Ecological Society., Journal of Ecology
Determinants of flammability in savanna grasses 9
van Altena, C., van Logtestijn, R.S.P., Cornwell, W.K. & Cornelissen, J.H.C.
(2012) Species composition and fire: non-additive mixture effects on ground
fuel flammability. Frontiers in Plant Science,3, 63.
Anderson, H.E. (1970) Forest fuel ignitability. Fire Technology,6, 312–319.
Anderson, H.E. (1982). Aids to determining fuel models for estimating fire
behavior. USDA Forest Service, Intermountain Forest and Range Experiment
Station. General Technical Report INT-122, 22.
Archibald, S., Lehmann, C.E.R., Gomez-Dans, J.L. & Bradstock, R.A. (2013)
Defining pyromes and global syndromes of fire regimes. Proceedings of the
National Academy of Sciences of the United States of America,110, 6442–
6447.
Beckage, B., Platt, W.J. & Gross, L.J. (2009) Vegetation, fire, and feedbacks: a
disturbance mediated model of savannas. The American Naturalist,174,
805–818.
Belcher, C.M., Mander, L., Rein, G., Jervis, F.X., Haworth, M., Hesselbo, S.P.,
Glasspool, I.J. & McElwain, J.C. (2010) Increased fire activity at the Trias-
sic/Jurassic boundary in Greenland due to climate-driven floral change. Nat-
ure Geoscience,3, 426–429.
Benson, D.A., Karsch-Mizrachi, I., Clark, K., Lipman, D.J., Ostell, J. & Sayers,
E.W. (2012) GenBank. Nucleic Acids Research,40, D48–D53.
Boettiger, C., Coop, G. & Ralph, P. (2012) Is your phylogeny informative?
Measuring the power of comparative methods. Evolution,66, 2240–2251.
Bond, W.J. (2008) What limits trees in C
4
grasslands and savannas? Annual
Review of Ecology Evolution and Systematics,39, 641–659.
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. Population and Com-
munity Biology Series, 14. Chapman & Hall, London, UK.
Bond, W.J., Woodward, F.I. & Midgley, G.F. (2005) The global distribution of
ecosystems in a world without fire. New Phytologist,165, 525–537.
Bradstock, R.A. & Auld, T.D. (1995) Soil temperature during experimental
bushfire in relation to fire intensity: consequences for legume germination
and fire management in south-eastern Australia. Journal of Applied Ecology,
32,76–84.
Brooks, M.L., D’Antonio, C.M., Richardson, D.M., Grace, J.B., Keeley, J.E.,
DiTomaso, J.M., Hobbs, R.J., Pellant, M. & Pyke, D. (2004) Effects of inva-
sive alien plants on fire regimes. BioScience,54, 677–688.
Byram, G.M. (1959) Combustion of forest fuels. Forest Fire: Control and Use
(ed K.P. Davis), pp. 61–89. McGraw-Hill, New York, NY, USA.
Christin, P.-A., Spriggs, E., Osborne, C.P., Str€
omberg, C.A.E., Salamin, N. &
Edwards, E.J. (2014) Molecular dating, evolutionary rates, and the age of the
grasses. Systematic Biology,63, 153–165.
Cornwell, W.K., Elvira, A., van Kempen, L., van Logtestijn, R.S.P., Aproot, A.
& Cornelissen, J.H.C. (2015) Flammability across the gymnosperm phy-
logeny: the importance of litter particle size. New Phytologist,206, 672–681.
D’Antonio, C.M. & Vitousek, P.M. (1992) Biological invasions by exotic
grasses, the grass/fire cycle, and global change. Annual Review of Ecology
and Systematics,23,63–87.
Drummond, A.J. & Rambaut, A. (2007) BEAST: bayesian evolutionary analy-
sis by sampling trees. BMC Evolutionary Biology,7, 214.
Emerson, B.C. & Gillespie, R.G. (2008) Phylogenetic analysis of community
assembly and structure over space and time. Trends in Ecology & Evolution,
23, 619–630.
Everson, C.S., Everson, T.M. & Tainton, N.M. (1988) Effects of intensity and
height of shading on the tiller initiation of 6 grass species from the highland
sourveld of natal. South African Journal of Botany,54, 315–318.
Fernandes, P.M. & Cruz, M.G. (2012) Plant flammability experiments offer
limited insight into vegetation-fire dynamics interactions. New Phytologist,
194, 606–609.
Fonda, R.W. (2001) Burning characteristics of needles from eight pine species.
Forest Science,47, 390–396.
Forrestel, E.J., Donoghue, M.J. & Smith, M.D. (2014) Convergent phylogenetic
and functional responses to altered fire regimes in mesic savanna grasslands
of North America and South Africa. New Phytologist,203, 1000–1011.
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.
Ganteaume, A., Jappiot, M., Lampin, C., Guijarro, M. & Hernando, C. (2013)
Flammability of some ornamental species in wildland-urban interfaces in
southeastern France: laboratory assessment at particle level. Environmental
Management,52, 467–480.
Gill, A.M. & Moore, P.H.R. (1996) Ignitability of Leaves of Australian Plants,
p. 34. CSIRO Plant Industry, Canberra, Australia.
Gill, A.M. & Zylstra, P. (2005) Flammability of Australian forests. Australian
Forestry,68,87–93.
Glasspool, I.J., Edwards, D. & Axe, L. (2004) Charcoal in the Silurian as evi-
dence for the earliest wildfire. Geology,32, 381–383.
Grass Phylogeny Working Group II (2012) New grass phylogeny resolves deep
evolutionary relationships and discovers C
4
origins. New Phytologist,193,
304–312.
Grigulis, K., Lavorel, S., Davies, I.D., Dossantos, A., Lloret, F. & Montserrat,
V. (2005) Landscape-scale positive feedbacks between fire and expansion of
the large tussock grass, Ampelodesmos mauritanica in Catalan shrublands.
Global Change Biology,11, 1042–1053.
Grootemaat, S., Wright, I.J., van Bodegom, P.M., Cornelissen, J.H.C. &
Cornwell, W.K. (2015) Burn or rot: leaf traits explain why flammability and
decomposability are decoupled across species. Functional Ecology,29,
1486–1497.
Hadfield, J.D. (2010) MCMC methods for multi-response generalized linear
mixed models: the MCMCglmm R package. Journal of Statistical Software,
33,1–22.
He, T., Lamont, B.B. & Downes, K.S. (2011) Banksia born to burn. New Phy-
tologist,191, 184–196.
Hughes, F., Vitousek, P.M. & Tunison, T. (1991) Alien grass invasion and fire
in the seasonal submontane zone of Hawaii. Ecology,72, 743–746.
Keeley, J.E., Pausas, J.G., Rundel, P.W., Bond, W.J. & Bradstock, R.A. (2011)
Fire as an evolutionary pressure shaping plant traits. Trends in Plant Science,
16, 406–411.
Lamont, B.B., Le Maitre, D.C., Cowling, R.M. & Enright, N.J. (1991) Canopy
seed storage in woody plants. Botanical Review,57, 277–317.
de Magalh~
aes, R.M.Q. & Schwilk, D.W. (2012) Leaf traits and litter flammabil-
ity: evidence for non-additive mixture effects in a temperate forest. Journal
of Ecology,100, 1153–1163.
Martin, T. (2010) Photosynthetic and evolutionary determinants of the response
of selected C
3
and C
4
(NADP-ME) grasses to fire. MSc Thesis, Rhodes
University, Grahamstown, South Africa.
Milton, S.J. (2004) Grasses as invasive alien plants in South Africa. South Afri-
can Journal of Science,100,69–75.
Murray, B.R., Hardstaff, L.K. & Phillips, M.L. (2013) Differences in leaf
flammability, leaf traits and flammability-trait relationships between native
and exotic plant species of dry sclerophyll forest. PLoS One,8, e79205.
Nelson, R.M. Jr (2001) Water relations of forest fuels. Forest Fires: Behavior
and Ecological Effects (eds E.A. Johnson & K. Miyanishi), pp. 79–149. Aca-
demic Press, San Diego, CA, USA.
Orme, D., Freckleton, F.P., Thomas, G.H., Petzoldt, T., Fritz, S., Isaac, N. &
Pearse, W. (2012) The Caper package: comparative analysis of phylogenet-
ics and evolution in R. Available at: http://cran.r-project.org/web/packages/
caper.
Overbeck, G.E. & Pfadenhauer, J. (2007) Adaptive strategies in burned sub-
tropical grassland in southern Brazil. Flora,202,27–49.
Papio, C. & Trabaud, L. (1991) Comparative-study of the aerial structure of 5
shrubs of mediterranean shrublands. Forest Science,37, 146–159.
Pausas, J.G. & Bradstock, R.A. (2007) Fire persistence traits of plants along a
productivity and disturbance gradient in mediterranean shrublands of south-
east Australia. Global Ecology and Biogeography,16, 330–340.
Pausas, J.G. & Keeley, J.E. (2014) Evolutionary ecology of resprouting and
seeding in fire-prone ecosystems. New Phytologist,204,55–65.
Pausas, J.G., Alessio, G.A., Moreira, B. & Corcobado, G. (2012) Fires enhance
flammability in Ulex parviflorus.New Phytologist,193,18–23.
Philpot, C.W. (1969) Seasonal changes in heat content and ether extractive
content of chamise. Research Paper INT-61. USDA Forest Service, Inter-
mountain Forest and Range Experiment Station, Ogden, UT, USA.
Plucinski, M.P. & Anderson, W.R. (2008) Laboratory determination of factors
influencing successful point ignition in the litter layer of shrubland vegeta-
tion. International Journal of Wildland Fire,17, 628–637.
Pyne, S.J. (1984) Introduction to Wildland Fire –Fire Management in the Uni-
ted States. Wiley, New York, NY, USA.
R Core Team (2013) R: A Language and Environment for Statistical Comput-
ing. R Foundation for Statistical Computing, Vienna, Austria. Available at
http://www.R-project.org/.
Ripley, B., Donald, G., Osborne, C.P., Abraham, T. & Martin, T. (2010)
Experimental investigation of fire ecology in the C
3
and C
4
subspecies of
Alloteropsis semialata.Journal of Ecology,98, 1196–1203.
Ripley, B., Visser, V., Christin, P.-A., Archibald, S., Martin, T. & Osborne,
C.P. (2015) Fire ecology of C
3
and C
4
grasses depends on evolutionary
history and frequency of burning but not photosynthetic type. Ecology,96,
2679–2691.
Rossiter, N.A., Setterfield, S.A., Douglas, M.M. & Hutley, L.B. (2003) Testing
the grass-fire cycle: alien grass invasion in the tropical savannas of northern
Australia. Diversity and Distribution,9, 169–176.
©2015 The Authors. Journal of Ecology published by John Wiley &Sons Ltd on behalf of British Ecological Society., Journal of Ecology
10 K. J. Simpson et al.
Rothermel, R.C. (1972) A mathematical model for predicting fire spread in
wildland fuels. Research Paper INT-115. USDA Forest Service, Intermoun-
tain Forest and Range Experiment Station, Ogden, UT, USA.
Saura-Mas, S., Paula, S., Pausas, J.G. & Lloret, F. (2010) Fuel loading and
flammability in the Mediterranean basin woody species with different post-fire
regenerative strategies. International Journal of Wildland Fire,19, 783–794.
Scarff, F.R. & Westoby, M. (2006) Leaf litter flammability in some semi-arid
Australian woodlands. Functional Ecology,20, 745–752.
Schwilk, D.W. (2003) Flammability is a niche construction trait: canopy archi-
tecture affects fire intensity. American Naturalist,162, 725–733.
Schwilk, D.W. & Caprio, A.C. (2011) Scaling from leaf traits to fire behaviour:
community composition predicts fire severity in a temperate forest. Journal
of Ecology,99, 970–980.
Silva, I.A. & Batalha, M.A. (2010) Phylogenetic structure of Brazilian savannas
under different fire regimes. Journal of Vegetation Science,21, 1003–1013.
Tansey, K., Gr
egoire, J.-M., Defourny, P., Leigh, R., Pekel, J.-F., van Bogaert,
E. & Bartholom
e, E. (2008) A new, global, multi-annual (2000–2007) burnt
area product at 1 km resolution. Geophysical Research Letters,35, L01401.
Tewarson, A. (2002) Generation of heat and chemical compounds in fires. The
SFPE Handbook of Fire Protection Engineering, 3rd edn (eds P.J. DiNenno,
D. Drysdale, C.L. Beyler & W.D. Walton), pp. 3–82. National Fire Protec-
tion Association, Quincy, MD, USA.
Trollope, W.S.W. (1978) Fire behaviour –a preliminary study. Proceedings of
the Grassland Society of South Africa,13, 123–128.
Trollope, W.S.W. (1984) Fire in savanna. Ecological Effects of Fire in South
African Ecosystems (eds V. Booysen & N.M. Tainton), pp. 200–217.
Springer-Verlag, Berlin, Germany.
Uys, R.G. (2000) The effects of different burning regimes on grassland phytodiver-
sity. MSc thesis, Botany Department, University of Cape Town, South Africa.
Uys, R.G., Bond, W.J. & Everson, T.M. (2004) The effect of different fire
regimes on plant diversity in southern African grasslands. Biological Conser-
vation,118, 489–499.
Vacchiano, G. & Ascoli, D. (2014) An implementation of the Rothermel fire
spread model in the R programming language. Fire Technology,50, 823–
1042.
Verd
u, M. & Pausas, J.G. (2007) Fire drives phylogenetic clustering in Mediter-
ranean Basin woody plant communities. Journal of Ecology,95, 1316–1323.
Visser, V., Woodward, F.I., Freckleton, R.P. & Osborne, C.P. (2012) Environ-
mental factors determining the phylogenetic structure of C
4
grass communi-
ties. Journal of Biogeography,39, 232–246.
Received 19 March 2015; accepted 26 October 2015
Handling Editor: Hans Cornelissen
Supporting Information
Additional Supporting Information may be found in the online ver-
sion of this article:
Figure S1. Schematic drawing of the set-up used to measure plant-
scale combustibility and sustainability.
Figure S2. Cumulative dry biomass over vertical plant height for the
grass species.
Figure S3. The influence of plant traits on components of Rother-
mel’s (1972) fire spread rate model.
Table S1. Climate data from plant collection sites.
Table S2. Grass species names, collection site and GenBank acces-
sion details.
Table S3. Plant traits values used to model the forward rate of fire
spread (m min
1
).
Table S4. Species mean flammability component values.
Table S5. Species mean plant trait values.
Table S6. Results of analysis of variance (two-way ANOVA with inter-
action) of leaf-scale flammability by species and state (fresh or dry).
Table S7. Mean plant trait values for the three collection sites.
©2015 The Authors. Journal of Ecology published by John Wiley &Sons Ltd on behalf of British Ecological Society., Journal of Ecology
Determinants of flammability in savanna grasses 11