- Top of page
Ibanez et al. (Journal of Vegetation Science, 2013, 24: 177) used fire modelling to examine the differential fire tolerance of secondary forest species in New Caledonia savanna. Misinterpretation of the concept of flame-front residence time and inadequate choices and linkages of fire behaviour and fire effects models affected the estimation of both stem and canopy injury. The most relevant concern is that the depth of fire-induced necrosis in the stem was overemphasized by a factor of two to three, underestimating tree tolerance to fire and overestimating the tree size thresholds for escaping fire.
Fire-inflicted tree damage and mortality reflect the interactions between fire behaviour characteristics and plant tissue insulation and distance from the heat source. Ibanez et al. (2013) considered these influences to model the consequences of exposing to fire a number of forest species growing in the savannas of New Caledonia. Vascular cambium and crown injuries were, respectively, described by the depth of bole necrosis (Dn), as measured from the bark surface, and by crown scorch height (HSC). The former was estimated from flame residence time (Rt) and fireline intensity (I) with an equation that Bova & Dickinson (2005) derived from experimental fires including heat flux measurements. Cambium necrosis occurs when Dn exceeds bark thickness.
The equation of Bova & Dickinson (2005) is empirical and describes cambium injury in Acer rubrum and Quercus prinus, ensuing the combustion of a mixture of deciduous hardwood litter and artificial fuels. This unique experimental setting clearly departs from the grass-dominated fuel complex studied by Ibanez et al. (2013). Additionally, Bova & Dickinson (2005) state, ‘we do not know how broadly our results can be applied to other tree species.’ By combining the duration of lethal stem heating (a function of burning time and fuel consumption) with critical time for cambium kill (a function of squared bark thickness), Peterson & Ryan (1986) offer a conceptually more robust framework to model cambium kill. However, neither of these modelling approaches is able to account for the stem damage effects of inter-specific variation in bark structure, density and moisture (Hengst & Dawson 1994; Brando et al. 2012).
In Bova & Dickinson (2005), Rt equals the ‘time over which flames are present at a given location on the surface of the bole,’ which they assume to equal the flame front residence time, the duration of flaming combustion at a given location. Because fire spreads as a band of finite width (the flame depth), it follows that Rt equals flame depth divided by the rate of fire spread (Rothermel & Deeming 1980; Alexander 1982). Apparently, Ibanez et al. (2013) misunderstood the meaning of flame-front residence time by assuming Rt = 60 s because it is representative of the persistence of temperatures above 60 °C measured in West African savanna woodland experimental fires (Savadogo et al. 2007). Plant tissues that are heated to temperatures in the order of 60 °C are assumed to be killed by fire (Peterson & Ryan 1986). However, fire temperatures and the temperatures effectively sustained by the stem cambium are unrelated. Note that passage of the fire-front is associated with flame temperatures >300 °C, as measured in field experiments (Taylor et al. 2004; Wotton et al. 2012). Flaming residence times in the order of 30–60 s are typical of forest litter (e.g. Taylor et al. 2004; Wotton et al. 2012). The experimental grass fires data set of Sneeuwjagt & Frandsen (1977) indicates that, on average, Rt = 8.5 s (range: 2.1–25.7 s) and Cheney & Sullivan (2008) characterize grassy fuels has having flame residence times of 5–15 s. The influence of Rt in depth of necrosis estimation prevails over the influence of fireline intensity; hence its substantial overestimation by Ibanez et al. (2013) inflates Dn by a factor of two to three (Fig. 1). Depending on Rt, the fireline intensity causing stem death for a given Dn changes considerably. As an example, Dn = 5 mm corresponds to 50% stem survival (rule of thumb; Lawes et al. 2011a) when I ≈ 1000 kW·m−1 and Rt = 20 s and when I ≈ 10 000 kW·m−1 and Rt = 10 s.
Figure 1. Stem-tissue depth of necrosis variation with Byram's fireline intensity (100–15 000 kW·m−1) for flame-front residence times of 10, 20 and 60 s, as per Bova & Dickinson (2005).
Download figure to PowerPoint
Fireline intensity (kW·m−1) in Bova & Dickinson (2005) conforms to Byram (1959) and is computed as the product of rate of fire spread (m·s−1), available fuel consumption (kg·m−2) and the net low heat of combustion (kJ·kg−1). Ibanez et al. (2013) simulated fire behaviour with BehavePlus (Andrews et al. 2008), which implements the formulation of Albini (1976) for fireline intensity as I = IR. r. Rt, where IR and r stand for reaction intensity (kW·m−2) and rate of fire spread (m·s−1). Hence, Ibanez et al. (2013) could have determined Rt from Albini's equation. Although this procedure generally produces underestimates (Cruz & Alexander 2010), it is in reasonable agreement with field observations in grass fuels (Sneeuwjagt & Frandsen 1977). BehavePlus estimates of fireline intensity for the original 13 standard US fuel models are lower than those obtained through the original Byram's formulation by a factor of two to three (Cruz et al. 2004). Hence, Ibanez et al. (2013) linked fireline intensity and stem injury through a different method than Bova & Dickinson (2005), which is expected to underestimate Dn. To elude this shortcoming, Byram's fireline intensity could have been estimated by combining the fuel loadings in their Appendix S1 with the fire spread rates predicted either with BehavePlus or with an adequate empirical-based model, namely that of Cheney et al. (1998).
A fuel model can be built to assess the effect of modifying the fireline intensity formulation using Appendix S1 mean fuel depth and fuel loadings by fuel category and assuming Van Wilgen & Wills (1988) figures for the remaining inputs. As an example, a BehavePlus prediction of r = 0.117 m·s−1 corresponds to Albini's I = 2073 kW·m−1, Rt = 12.5 s and Dn = 4.9 mm. For the same fire spread rate and Rt, Byram's I = 3047 kW·m−1 and Dn = 5.3 mm, assuming 16 890 kJ·kg−1 for heat yield (Govender et al. 2006) and 1.546 kg·m−2 for fuel consumption, i.e. 90% of the total fuel load as observed in northern Australia savanna fires when I > 2000 kW·m−1 (Williams et al. 1998). The difference in Dn is modest, but note that BehavePlus estimates of fire characteristics are highly responsive to variations in fuel parameters. Grass surface-to-volume ratios in Van Wilgen & Wills (1988) are conservatively low; an increase to 9770 m−1 (Cheney et al. 1993) in the fuel model would imply that Byram's I would double in relation to Albini's I.
Ibanez et al. (2013) estimated the height of crown scorch (m) through HSC = k. I2/3, replacing the original k coefficient (0.148; Van Wagner 1973) with k = 0.059 as an approximation to the equation fitted by Williams et al. (1998) to savanna crown-scorch data. Again, the use of Albini's I to predict crown scorch height does not respect Van Wagner's formulation, leading to underestimation of both scorch height and crown fire likelihood, as thoroughly discussed in Alexander & Cruz (2012a). Still, the impact of varying how HSC is predicted is largely irrelevant for most of the fireline intensity range, given the low stature of trees on the study sites: the average canopy (6.9-m high) is fully scorched when I reaches 1208 or 306 kW·m−1, which, respectively, correspond to using k = 0.059 or k = 0.148.
The revised fire modelling results imply lower DBH thresholds for the avoidance of fire-induced stem necrosis than reported in Ibanez et al. (2013), i.e. lower escape size. Taking as an example the low tree Fagraea berteoroana and assuming a high 20-s Rt, these thresholds will dramatically decrease from 20–37 cm, depending on fireline intensity (fig. 3 in Ibanez et al. 2013), to 5–11 cm. Dn in high-intensity savanna fires (allowing for a 5000–20000 kW·m−1 variation in Byram's I) varies from 5.0 to 10.4 mm when Rt is within the 10–20 s interval. This does not imply that trees with thicker bark are free from localized or partial cambium necrosis, namely because the Bova & Dickinson (2005) equation reflects stem heating on its windward side; leeward Rt will be higher, especially for larger trees and under stronger winds (Gutsell & Johnson 1996).
New Caledonian early-successional forest species colonizing savanna are certainly more tolerant of stem heating than contended by Ibanez et al. (2013). Protecting the bole within the flaming zone from heat does not seem to warrant the existing large variation in bark thickness investment among species thriving in an environment characterized by short flaming time. This is consistent with alternative strategies to resist fire (Lawes et al. 2011b) but it also suggests that the role of bark thickness in top-kill avoidance is more related to ensuring bud survival at higher heights in the tree, which distinguishes crown scorch from crown kill. Taller trees will experience a more favourable time–temperature regime and will be less likely to actually burn (as opposed to be scorched), regardless of bark thickness. Judging from the canopy base height figures in Appendix S1, partial or total crown combustion is a real possibility in the studied savanna, presumably with more severe consequences on resprouting than crown scorch.
The results of correctly using the fire modelling approach adopted in Ibanez et al. (2013) rebut their conclusion that the degree of fire adaptation of early-successional forest species in New Caledonia is insufficient to assure that young individuals survive to maturity. Quoting from Van Wagner & Methven (1978), ‘these comments are not made in a spirit of negative criticism, but rather in the interest of promoting proper attention to fire behaviour on the part of ecologists studying fire effects.’ Poor understanding of the physical mechanisms involved is common in both experimental and simulation studies of fire behaviour (Alexander & Cruz 2012b; Fernandes & Cruz 2012). To avoid misleading results, simulation modelling in fire research applications should be well informed on the underlying concepts and model tenets, be aware of the feasibility of coupling distinct and independently developed models, and ensure that model inputs are realistic and representative.