Costs and benefits of relative bark thickness in relation to fire damage: a savanna/forest contrast

Authors


Correspondence author. E-mail: Michael.Lawes@cdu.edu.au

Summary

  1. In fire-prone ecosystems, bark protects the stem bud bank from fire. Absolute bark thickness is a good indicator of this protective function, but it depends on stem size as well as inherent differences between species. Relative bark thickness (i.e. relative to stem diameter) takes the latter into account. We argue that relative bark thickness is an important functional trait offering insights to the evolution of species persistence in fire-prone habitats.
  2. During growth ontogeny different species can acquire absolutely thick bark through having: (i) relatively thick bark (i.e. an early commitment to thick bark) or (ii) relatively thin bark but fast stem diameter growth rates. We test the hypothesis that the most effective way of protecting tree stems from frequent fire is by having relatively thick-barked small stems. We predict that species with higher relative bark thickness are more common in fire-prone habitats. In habitats with long fire-free intervals such as rainforest, delayed investment in bark thickness results in thin bark.
  3. We examined the relative bark thickness of woody congeners from Australian non-fire-prone forest and fire-prone savanna and in other tree-dominated systems world-wide. We determined the relative cost of acquiring absolute bark thickness of 0.5 cm for different rates of bark allocation. The insulating benefits of bark were considered a linear function of bark thickness.
  4. Synthesis. We suggest that relatively thick bark minimizes the costs of acquiring absolutely thick bark, and it confers greater protection to smaller stems. The cost of acquiring thick bark prevents small trees from merely accumulating bark as a consequence of fast height or stem diameter growth. Accordingly, our field survey indicated that forest species had relatively thin bark and acquired thick bark only as a consequence of very large size, while fire-prone savanna species had relatively thick-barked small stems. Based on this, relative bark thickness appears to be a good predictor of local fire regimes and is a useful plant functional trait.

Introduction

The fire resistance properties of trees, particularly their absolute bark thickness at the time of fire, determine species persistence, the capacity for resprouting (Hoffmann et al. 2009, 2012; Lawes & Clarke 2011; Lawes et al. 2011a,b), and even tree community composition (Lawes et al. 2011a). Comparative analyses of the functional significance of bark thickness have, however, stalled because there is (i) no uniform agreement on what constitutes the variable ‘bark thickness’, (ii) there have been only limited theoretical analyses of bark thickness (Jackson, Adams & Jackson 1999), and (iii) there are insufficient phylogenetically independently controlled data available on bark thickness (Hoffmann, Orthen & Do Nascimento 2003). In this paper, we challenge the usefulness of measuring absolute bark thickness as a functional trait. We argue that relative bark thickness is a more useful indicator of adaptive response to disturbance. We test the latter proposition by comparing the variation in bark thickness between fire-prone savanna and non-fire-prone forest in northern Australia and to other biomes.

Several studies have shown that trees ‘escape’ the effects of ground fires by protecting their stem with thick bark (Gignoux, Clobert & Menaut 1997; Hegde, Chandran & Gadgil 1998; Michaletz & Johnson 2007; Nefabas & Gambiza 2007; Lawes et al. 2011a,b). The premise of this argument is that thick bark is an adaptive trait (Keeley et al. 2011), which functions to protect the stem from heat damage (Lawes et al. 2011a,b). As an adaptive trait, bark thickness is potentially an easily measured key plant attribute that may offer insights to plant evolutionary and disturbance history (Keeley et al. 2011). However, the measurement of the attributes of bark thickness differs among researchers: maximum bark thickness (Keeley & Zedler 1998), absolute bark thickness (Cornelissen et al. 2003), allometric coefficient of log–log relationship between bark thickness and stem diameter (Jackson, Adams & Jackson 1999) and various measures of relative bark thickness (Hoffmann, Orthen & Do Nascimento 2003; Hoffmann et al. 2009; Lawes et al. 2011a). Relative bark thickness (i.e. percentage of the stem that is bark; Hoffmann, Orthen & Do Nascimento 2003) is a measure of individual bark thickness allocation by stem diameter (Hoffmann et al. 2009) and can be construed as the relative allocation to bark. The use of relative bark thickness has been refined in some studies to those stem size classes most susceptible to the effects of fire, that is, juvenile stems (Lawes et al. 2011a).

Hoffmann and colleagues (Hoffmann, Orthen & Do Nascimento 2003; Hoffmann & Solbrig 2003; Hoffmann et al. 2009) contend that epicormic resprouting plants in fire-prone habitats, but especially savannas, acquire absolutely thick bark in three ways: (i) by an early high rate of resource allocation to bark per stem diameter (i.e. relatively thick bark); (ii) as a natural consequence of fast height and thus diameter growth rate at more productive sites – in this way, relatively thin-barked species may acquire sufficiently thick bark at a critical size or age to persist through fire, essentially out-growing fire effects and/or by (iii) bark accumulation in relatively long inter-fire periods. Site productivity and annual bark growth rates are more difficult to determine than absolute or relative bark thickness. Absolute bark thickness is not a species-specific trait because it is dependent on the size of the individual being measured. In contrast, relative bark thickness is easy to measure and as it is the ratio of bark thickness to stem diameter it is less tree size dependent (Hoffmann et al. 2009) and is therefore a potentially desirable and useful bark trait.

The hypothesis that trees with relatively thin bark can achieve absolutely thick bark by growing rapidly can be tested by considering the costs and benefits of different levels of bark thickness relative to tree size (diameter). Because bark is the outer edge of the circumference of a circle, the resource allocation cost of bark should increase with increasing stem diameter. These costs may be modelled as proportional to the cross-sectional area of bark (Fig. 1). The benefits of bark in protecting the vascular cambium from fire effects are proportional to the (bark thickness)x, where usually  2 (van Mantgem & Schwartz 2003; Michaletz & Johnson 2007; Lawes et al. 2011b). We test the prediction that natural selection should favour relatively thicker bark under increasing fire disturbance severity.

Figure 1.

Diagram illustrating the relationship between absolute and relative bark thickness. Here, the bark is modelled as the outer edge or annulus about the circumference of a circle. The costs of bark are directly proportional to the cross-sectional area of bark and thus increase with increasing stem size for a given absolute bark thickness. Absolute bark thickness is dependent on the absolute size of the individual being measured, while for a given bark thickness (in this case 0.5 cm), relative bark thickness ((bark thickness)/(tree diameter) × 100) declines with tree size. Relative bark thickness is a useful bark thickness trait because it scales with tree size in an interpretable way. Relatively thicker bark at the juvenile stage is particularly informative as a potentially fire-adapted trait.

Several studies have demonstrated the importance of thick bark to the fire resilience of trees in savanna (Gignoux, Clobert & Menaut 1997; Gashaw et al. 2002; Hoffmann, Orthen & Do Nascimento 2003; Nefabas & Gambiza 2007; Lawes et al. 2011a; Hoffmann et al. 2012) and dry (Pinard & Huffman 1997), gallery (VanderWeide & Hartnett 2011) and disturbed tropical rainforest trees (Hegde, Chandran & Gadgil 1998; Barlow, Lagan & Peres 2003; Van Nieuwstadt & Sheil 2005; Paine et al. 2010). However, few studies other than Hoffmann and Franco ( 2003) and Hoffmann et al.(2009) have compared bark thickness among savanna and forest congeners. Here, we test whether trees of highly fire-prone tropical savannas (high rainfall, low soil fertility) have thicker bark than confamilial and congeneric tree species in monsoon forest (vine thicket) in northern Australia. By contrasting fire-prone north Australian savannas with other savannas and forests world-wide, we present insights into how allocation to thicker bark in fire-prone savanna tree species has evolved and thus the utility of relative bark thickness in predicting species persistence in fire-prone habitats.

Materials and methods

Field Measurements

The bark thickness of 43 tree species from 1020 individuals in tropical savanna, and 20 species from 281 individuals in monsoon forest, was sampled at Litchfield National Park (LNP; 1464 km2) in June 2009 and at East Point Reserve (135 ha) in August 2009, respectively. The vegetation in LNP is dominated by eucalypt open forests and woodlands, with a grassy understorey, referred to hereafter as tropical savanna. The seasonal climate is particularly conducive to high fire frequencies: the wet season is highly productive with abundant grass growth, while the 7-month dry season (c. May–November) strongly promotes grass curing and flammability. Fires are typically ground fires and are fuelled by native Sorghum spp. grasses. On average 66% of LNP is burnt per annum (Russell-Smith et al. 2009). Data from tropical savannas were collected from vegetation monitoring plots (20 m × 40 m) that were established between March 1994 and August 1996 in LNP. These monitoring plot data indicate that 86% of fires in LNP are of low (2 m scorch height) or moderate (sub-canopy scorch) severity. We sampled only those plots that had been burnt at least in the previous dry season or more recently and had been burnt in at least five of the previous 12 years, that is an average fire return interval of c. 1.5 years (Northern Territory Bushfires Council, unpublished data). Fire severity in the plots we surveyed was of low to moderate intensity.

East Point Reserve (12°24′30″ S, 130°50′ E) is a regionally significant remnant of coastal monsoon forest (Panton 1993). The forest belongs to Group 9 of Russell-Smith (1991) and is described as semi-deciduous rainforest and vine thicket. Monsoon forest on the East Point peninsula was extensively cleared and burnt from 1932 to 1963 (Franklin, Matthews & Lawes 2010), and we sampled only forest remnants that had not been previously cleared or burnt.

Bark thickness was measured using a standard thickness gauge (Haglöf, Barktax, Sweden) or where bark was not penetrable using the gauge, we used a needle punch (diam. = 3 mm). In both cases, the gauge was inserted to the point of resistance by the sapwood. Bark thickness measured by this method includes the cambium in most instances, but as the cambium is thin (< 0.5 mm), this did not substantially affect measures of bark thickness. Where bark was fissured, we measured thickness from the highest point. We took three evenly spaced measures of bark thickness around the circumference of the stem at breast height (1.3 m) per individual. The diameter at breast height of each tree was also measured.

Statistical Analyses

To control the effect of phylogeny on bark characteristics, the relative bark thickness of tree species from tropical savanna and monsoon forest were compared at two taxonomic levels: (i) matched in eight confamilial pairs and (ii) five congeneric pairs (Table 1). Analysis of variance based on an unbalanced design was used to analyse the effect of vegetation type (forest or savanna) on bark thickness for the taxonomic level (family, genus comparisons). Because bark thickness varies with tree size (Werner & Murphy 2001; Lawes et al. 2011a), relative bark thickness (the ratio of individual bark thickness to stem diameter) was used as the response variable. Relative bark thickness was square root transformed. To compare absolute bark thickness without the confounding influence of stem diameter, in a separate analysis, we included stem diameter as a covariate in the model. Analyses were conducted using GenStat 13.3 (VSN-International 2011).

Table 1. Number of species surveyed at the family and genus level in this study
Taxonomic levelNo. of species
ForestSavanna
Family
Anacardiaceae12
Ceasalpiniaceae21
Combretaceae13
Euphorbiaceae21
Mimosaceae11
Rubiaceae12
Sapotaceae11
Sterculiaceae11
Genus
Acacia 12
Buchanania 11
Erythrophleum 11
Pouteria 11
Terminalia 13
Species
Erythrophleum chlorostachys 11

The allometric coefficient (slope) for the relationship between bark thickness and stem diameter is variously referred to as the allocation coefficient, the bark coefficient, the relative bark thickness slope or the bark ratio (Jackson, Adams & Jackson 1999; Hoffmann et al. 2009). The bark ratio (b) is the same as relative bark thickness per se and has been applied by some as a population- or community-wide relationship as a habitat bark ratio (Hoffmann et al. 2009), where the individual relative bark thicknesses are averaged across a large sample of species. Relative bark thickness or the bark ratio should ideally be determined at those small size classes where fire most affects stem survival. We fitted the bark ratio based on a linear regression equation to our own data from northern Australia to be consistent with comparative data from other sites world-wide (Hoffmann et al. 2009). Accordingly, regression fits were forced through the origin. For graphic comparison of the matched confamilial or congeneric pairs between habitats in north Australia, we fitted the logarithmic function to the data (BT = a + b·Loge D; where BT = bark thickness, and = tree diameter), as this better describes the asymptotic nature of bark thickness with increasing tree diameter. The comparison of the linear bark ratio among several forest and woodland sites world-wide, including our own data from tropical savannas, is not intended to be an exhaustive analysis of fire-prone biomes, and data were selected from Cerrado (Hoffmann et al. 2009) and Amazonian rainforest (Uhl & Kauffman 1990) only. We used the program xyExtract Graph Digitizer 5.1 (Wilton and Cleide Pereira da Silva, Campina Grande, Brazil; http://zeus.df.ufcg.edu.br/labfit/index_xyExtract.htm) to extract a reliable facsimile of bark thickness and stem diameter data from figures in Hoffmann et al. (2009) and Uhl & Kauffman (1990).

The Costs and Benefits of Thick Bark

The cost to thin- vs. thick-barked species of accumulating bark was modelled based on the change in cross-sectional area of bark with increasing stem diameter growth (Fig. 1). We used a threshold bark thickness of 0.5 cm (50% stem survival after fire) to calculate the relative costs of bark growth among plants with different relative bark thickness. Hoffmann et al. (2009) found that a bark thickness of 0.62 cm was required for 50% stem survival after fire in Cerrado, while Lawes et al. (2011a) estimated threshold bark thickness for 50% stem survival at 0.25 cm for eucalypt and 0.35 cm for non-eucalypt species in Australian tropical savannas, and a minimum bark thickness of 0.4–0.5 cm for eucalypt species to resist the effects of repeated fire. Thus, the threshold value of bark thickness of 0.5 cm is roughly the mean value for estimates provided by Hoffmann et al. (2009) and Lawes et al. (2011a).

Assuming that the cost of thick bark is a function of the cross-sectional area of bark, then this cost is given by:

display math(eqn 1)

where BA, cross-sectional area of the bark (cm2); D, diameter of the stem (cm); BT, bark thickness (cm).

Eqn (eqn 1) solves to:

display math(eqn 2)

So that, for a given BT, the BA is linearly proportional to the diameter of the stem (e.g. for BT = 0.5 cm, BA = 1.5708D − 0.7854; Fig. 1). To account for the habitat-specific rate of increase in BA, bark thickness (BT) in eqn (eqn 1) is derived from D × b, where b is the habitat bark ratio, so that the increase in BA for a habitat is given by:

display math(eqn 3)

The benefit of thick bark is modelled as a function of the insulating or heating properties of the bark (Peterson & Ryan 1986; Lawes et al. 2011b), specifically the inverse of the duration of heating required to kill the cambium of a tree (τc). The duration of heating required to kill the cambium of a tree (τc) is directly proportional to bark thickness squared (Peterson & Ryan 1986; van Mantgem & Schwartz 2003; Lawes et al. 2011b). This relationship (τc), based on a fire with a constant temperature and α = 0.060 cm2 min−1, where α is the bark thermal diffusivity or rate (cm2 min−1) at which heat is transferred through a given thickness of bark (BT) from the bark surface to the cambium (Peterson & Ryan 1986), is a measure of the insulation that bark provides to a tree stem. Peterson & Ryan (1986) found this relationship for conifers to be:

display math(eqn 4)

While for trees in the tropical savannas of northern Australia, Lawes et al. (2011b) report:

display math(eqn 5)

Here, we use the general relationship inline image to model the benefit of thick bark. To derive the habitat-specific relationship, bark thickness is calculated as the product of tree diameter (D) and the habitat-specific bark ratio (b). Thus, the benefit of thicker bark (1/τc) for a habitat is given as:

display math(eqn 6)

Results

Combining the data for the eight families compared in this study, on average absolute bark thickness of savanna trees (mean ± SE = 9.7 ± 0.22 mm, n = 276) was about double that of forest trees (mean ± SE = 5.2 ± 0.33 mm, n = 122; F1,395 = 155.1, < 0.001). Relative bark thickness was greater in savanna than forest trees when compared at the family (F7,382 = 8.2, < 0.001; Fig. 2) and the genus (F4,306 = 10.6, < 0.001) levels (Table 2), although these trends were less obvious in the field for the Mimosaceae or Acacia species, respectively (Fig. 2).

Table 2. Comparison of mean absolute bark thickness (see Materials and methods) and relative bark thickness among confamilial and congeneric pairs between monsoon forest and tropical savanna habitat. Relative bark thickness is the mean of the ratio of individual bark thickness to stem diameter (mm cm−1)
Taxonomic levelAbsolute bark thickness (mm) (mean ± SE) Relative bark thickness % = ratio × 10 (mean ± SE)
ForestSavannaForestSavanna
Family
 Anacardiaceae7.1 ± 0.110.9 ± 0.0040.82 ± 0.251.95 ± 0.1
 Ceasalpiniaceae6.2 ± 0.0210.0 ± 0.0030.37 ± 0.021.42 ± 0.08
 Combretaceae6.7 ± 0.018.4 ± 0.0040.52 ± 0.031.56 ± 0.1
 Euphorbiaceae3.3 ± 0.019.7 ± 0.010.65 ± 0.071.24 ± 0.05
 Mimosaceae6.0 ± 0.028.4 ± 0.020.41 ± 0.041.04 ± 0.1
 Rubiaceae4.7 ± 0.015.3 ± 0.060.59 ± 0.050.62 ± 0.09
 Sapotaceae5.4 ± 0.0114.6 ± 0.20.49 ± 0.031.81 ± 0.02
 Sterculiaceae3.4 ± 0.318.9 ± 0.30.291.71
Genus
 Acacia 6.3 ± 0.038.5 ± 0.010.41 ± 0.041.04 ± 0.1
 Buchanania 7.2 ± 0.211.1 ± 0.0050.82 ± 0.251.97 ± 0.1
 Erythrophleum 8.9 ± 0.210.1 ± 0.0040.29 ± 0.011.42 ± 0.08
 Pouteria 5.5 ± 0.0114.6 ± 0.20.49 ± 0.031.81 ± 0.02
 Terminalia 6.8 ± 0.018.5 ± 0.0050.52 ± 0.031.57 ± 0.1
Figure 2.

Relationship between stem diameter and bark thickness of confamilial savanna and forest trees. The slopes of the regression lines describe the relative bark thickness for a family.

When examined over all species including all individuals (n = 1289 individuals), tropical savanna species (dominated by eucalypts) had considerably thicker bark (higher habitat bark ratio) than monsoon forest species from northern Australia and Amazon forest species (Uhl & Kauffman 1990), similar bark thickness to gallery forest species from the Brazilian Cerrado (Hoffmann et al. 2009), but thinner bark than savanna species in fire-prone Cerrado (Fig. 3).

Figure 3.

Relationship between bark thickness and stem diameter in Australian tropical savanna and monsoon rainforest (this study; solid regression lines) compared to savanna and gallery forest in cerrado (Hoffmann et al. 2009) and Amazonian rainforest (Uhl & Kauffman 1990) (dashed regression lines). Australian tropical savanna stems are divided into eucalypt dominated (escape height) and non-eucalypt dominated (escape diameter) components after Lawes et al. (2011a). The slopes of the regression lines (bark ratio) are given in parentheses on the figure and are the bark allocation coefficients for each habitat.

Because bark area scales as an exponential function of bark thickness and as a linear function of tree diameter for a given bark thickness (coefficient of 1.57; see slope of grey line in Figs 1 and 4), relatively thin-barked trees need far greater cross-sectional bark area to achieve a given bark thickness (in our case 0.5 cm; Fig. 4). The net effect is that species with relatively thin bark incur greater allocation costs to achieve a threshold bark thickness for protection from heat damage. This prevents them acquiring thick bark by having high growth rates. Thinner barked species only acquired threshold bark thickness, if at all, at well beyond sapling size (Fig. 4) and are thus unable to grow sufficiently thick bark as saplings to ‘out-grow’ fire effects.

Figure 4.

Relationship between stem diameter and bark area compared between different habitats (and associated bark ratios). The grey line shows how the bark area, at a threshold bark thickness (taken as 0.5 cm), varies with relative bark thickness (habitat-specific bark ratio; Fig. 3) across stem diameter. Bark area represents the allocation or cost to achieving a specific bark thickness. This graphic model shows that the costs of bark thickness increase with decreasing bark ratio, so that thinner barked species achieve threshold bark thickness only well beyond sapling size.

While relatively thin-barked rainforest species do not acquire absolutely thick bark at the vulnerable juvenile stage, this appears possible for savanna species in general (see higher habitat bark ratio; Fig. 3), and for some fire-exposed forest types such as sclerophyll forests, gallery or monsoon forest patches (Fig. 3). Finally, the benefit of relatively thicker bark, in terms of protecting the vascular cambium and epicormic buds from heating (prolonged > 60 s exposure of the cambium to temperatures > 50–60 °C), is achieved at smaller stem diameters (> 2 cm diameter) for savanna than forest trees (> 5 cm diameter; Fig. 5).

Figure 5.

Relationship between stem diameter and the benefit of thicker bark for protecting the cambium and epicormic buds from fire. A rule of thumb is that prolonged exposure of the cambium to temperatures reaching 50–60 °C is sufficient to cause cambium necrosis. The benefit of thick bark is modelled as 1/τc – where higher values of 1/τc represent shorter heating times required to kill the cambium (e.g. a value of 1 is 60 s, while a value of 2 is 30 s of heating required to kill the cambium). The benefit of relatively thicker bark in savanna trees is achieved at smaller stem diameters than in forest and at sapling sized stems in savanna that are most vulnerable to fires.

Discussion

We predicted relatively thick bark in fire-dominated systems and our field observations (Lawes et al. 2011a,b), as well as those from other fire-prone systems (Hoffmann et al. 2009), support this generalization. Bark is relatively thicker in congeneric savanna than forest species in north Australia, and in general, this trend of relatively thicker bark in fire-prone than fire-absent habitats such as forest is observed world-wide (Fig. 3). These patterns are explicable in terms of the costs and benefits of thick bark in ecosystems subjected to varying levels of disturbance by fire. Because the fire protection benefits of bark do not scale linearly with bark thickness and larger stems may suffer relatively greater fire intensity (due to leeward eddies; Gill 1974; Michaletz & Johnson 2007), and because the costs of thick bark required to protect saplings from fire are traded off against growth at this crucial life stage (Lawes et al. 2011a), plants cannot escape the effects of fire by rapid height growth alone. In fact, we argue that it is unlikely any plants can out-grow the effects of frequent fire without protecting living tissue from heat damage and securing the capacity to both survive and recover after fire (Lawes & Clarke 2011). In other words, in fire-prone systems, it does not matter how fast or tall a tree sapling grows if it does not protect the stem from heat damage. Stems of relatively thin-barked individuals will likely die when fire intervals are shorter than the time required to achieve threshold bark thickness and selection for relatively thicker barked saplings is thus strong. Our data support the notion that thick bark in savanna trees is an adaptive trait (Keeley et al. 2011) and functions to protect the stem from heat damage (Lawes et al. 2011a,b). This protection function is not required in forests, where fire is not a selection pressure, and thus forest saplings have relatively thin bark as reflected in lower bark ratios in forested habitats when compared with savannas. We further contend that relative bark thickness is a useful key plant attribute reflecting plant evolutionary response in fire-prone ecosystems.

Among the suite of bark attributes, absolute bark thickness is most responsible for conferring fire resistance on an individual (Lawes et al. 2011b; and references therein). It is an easy trait to measure, it relates to the capacity to resprout and recover after fire (Clarke, Lawes & Midgley 2010; Lawes & Clarke 2011), and it has been widely recognized as a key dimension of plant ecological strategy in fire-prone ecosystems (Gignoux, Clobert & Menaut 1997; Hoffmann, Orthen & Do Nascimento 2003; Hoffmann & Solbrig 2003; Hoffmann et al. 2009; Midgley, Lawes & Chamaillé-Jammes 2010; Keeley et al. 2011; Lawes et al. 2011a,b). However, the use of relative bark thickness as a metric has not been thoroughly explored (Hoffmann et al. 2009). The application of relative bark thickness could be complicated by the fact that bark thickness may vary intraspecifically with stem size. For example, Jackson, Adams & Jackson (1999) showed that allometric analyses of bark thickness and stem diameter produced different bark ratios for a range of Pinus and Quercus species. Fortunately, Jackson, Adams & Jackson (1999) also showed that allometric coefficients correlated with bark thickness at a standardized size of 5 cm stem diameter. These smaller stem diameter classes are the most important size classes for where bark thickness may limit fire survival. Therefore, the measurement and application of relative bark thickness should be focussed on smaller stems and avoid large stems (> 30 cm diameter), which will reduce the need for allometric considerations.

Within taxonomic groups (e.g. eucalypts), relative bark thickness has been correlated with specialized resprouting (epicormic) anatomy (Burrows 2002; Burrows et al. 2010). Relatively thicker bark has undoubtedly evolved as a fire adaptation, and thus, the functional role of thicker bark must date back to the origins of fire as an ecosystem driver. Recent research has found a conserved phylogenetic relationship between post-fire resprouting (epicormic) anatomy and the evolution of the Myrtaceae-dominated Australian savannas dating from 60 to 62 Ma in the earliest Palaeogene (Crisp et al. 2011), some 50 million years earlier than previously thought (Beerling & Osborne 2006). Nutrient-poor Australian savannas are inherently fire-prone because they have intense sclerophylly (hard leaves) and associated long leaf retention times, which with low levels of herbivory make for very flammable conditions (Midgley & Bond 2011). In addition, low productivity in Australian savannas means that trees lack the potential for growth sufficient to escape fire effects and must protect their meristem and axillary buds from heat damage by having thick bark (Burrows et al. 2008; Lawes et al. 2011a) to ensure survival and recovery after fire. Thus, given the pervasive impact of fire on the Australian landscape and nearly all its biomes, it is notable that Australian monsoon forest species nevertheless have significantly thinner bark than their savanna counterparts, but also that their bark is in turn much thicker than for example the bark of Amazonian rainforest trees (Fig. 3). We offer an explanation for these trends below.

All evidence points to Australia being the home of fire with the longest history of fire–plant interactions (Midgley & Bond 2011), and thus, we might expect bark ratios to be highest in Australian fire-prone biomes; they are not (cf. Fig. 3). However, this overlooks the fact that lineages in Australian biomes, particularly the savanna biome, are also the most highly evolved in terms of other plant-fire adaptations (Burrows et al. 2010; Crisp et al. 2011; Keeley et al. 2011). The latter, and especially the uniquely deeply buried epicormic resprouting strand anatomy of the Myrtaceae (Burrows et al. 2010), suggests that over time, plants in Australian savannas have optimized their bark thickness (i.e. deeply buried meristematic strands can be protected by thinner bark), balancing the costs of producing and protecting their buds, to ensure that resprouting and the ability to persist through fire is not compromised (Lawes et al. 2011a,b). Their unique meristem strand anatomy offers an explanation for why eucalypt bark is relatively thinner in Australian savannas compared to other presumably more recently evolved fire-prone systems such as the cerrado. In these more recently evolved fire-prone systems, higher bark ratios may indicate the strength of selection for fire resistance traits because fire effects cause a strong binary response (i.e. survive or die) in the early stages of evolutionary response and adaptation.

Because fire residence time and thus fire intensity is positively related to stem size (Gill 1974), it is unlikely that plants can ‘out-grow’ fire effects and acquire sufficient bark thickness at the juvenile stage by fast stem growth alone. In fire-prone systems, allocation of reserves to bark growth, or a mechanism of sufficiently protecting buds, is a necessary co-requisite for persistence through fire (Gignoux, Clobert & Menaut 1997; Burrows 2002; Lawes et al. 2011a). Our conceptual model (Fig. 4) nevertheless predicts that species that are not fire-prone should have relatively thin bark, and fire-prone species should have relatively thicker bark. These predictions are by-and-large supported by comparisons of the bark ratio across habitats (Fig. 3). Relative bark thickness should therefore be a key predictor of local and historical fire regimes.

Acknowledgements

We thank Hylton Adie, Veronica French, Claire Haysum and Florence Durillon for assisting with collecting data in the field. We are grateful to John McCartney and the ranger staff of LNP for their support. This research was conducted under Permit Number 33520 issued by the Parks and Wildlife Commission of the Northern Territory. We are grateful to Roger Matthews and the Darwin City Council for permission to conduct forest research in East Point Nature Reserve.

Ancillary