Decreases in standing tree-based carbon stocks associated with repeated prescribed fires in a temperate mixed-species eucalypt forest

Forest Ecology and Management 306 (2013) 243–255

Contents lists available at SciVerse ScienceDirect

Forest Ecology and Management journal homepage: www.elsevier.com/locate/foreco

Full length article

Decreases in standing tree-based carbon stocks associated with repeated prescribed fires in a temperate mixed-species eucalypt forest Lauren T. Bennett a,⇑, Cristina Aponte b, Kevin G. Tolhurst a, Markus Löw a, Thomas G. Baker b a b

Department of Forest and Ecosystem Science, Melbourne School of Land and Environment, The University of Melbourne, 4 Water Street, Creswick, Victoria 3363, Australia Department of Forest and Ecosystem Science, Melbourne School of Land and Environment, The University of Melbourne, 500 Yarra Boulevard, Richmond, Victoria 3121, Australia

a r t i c l e

i n f o

Article history: Received 25 March 2013 Received in revised form 3 June 2013 Accepted 16 June 2013 Available online 24 July 2013 Keywords: Prescribed fire Planned burn Eucalypt forest Carbon Tree growth Tree mortality

a b s t r a c t Prescribed fire is a common management practice in fire-tolerant forests, and one that has potential carbon costs. Previous assessments of the carbon costs of prescribed fire regimes in temperate Australia have been based on little empirical data, and have focused on direct fire effects (area burnt, fuel consumed) but have largely ignored potentially substantive indirect effects on tree mortality and growth. This study measures effects of four prescribed fire treatments on standing tree-based carbon stocks, and on individual tree growth and mortality, in a fire-tolerant eucalypt forest of south-eastern Australia. Prescribed fire treatments were as a factorial combination of two seasons (autumn or spring) and two frequencies (3yearly ‘High’, or 10-yearly ‘Low’), were replicated over five study areas, and involved 2–7 low-intensity fires over 27 years. Total standing tree-based carbon stocks (live and dead) were significantly less in prescribed fire than control treatments. However, the mean carbon difference (25 Mg ha1) had a wide 95% confidence interval (2–48 Mg ha1), indicating a high degree of uncertainty about the magnitude of prescribed fire effects in these native forests. Overall decreases were consistent with detection of both direct and indirect effects of prescribed fire treatments. Direct combustion effects on bark were minimal (c. 0.2– 0.4 Mg ha1), but were also indicated by significantly less carbon in dead large stems in fire than control treatments despite evidence of marginally increased mortality of individual large stems in the former. Indirect effects of repeated prescribed fires were also detected as significantly decreased mean annual diameter increment of individual large Eucalyptus obliqua over 27 years (particularly of stems 20– 50 cm diameter). With respect to prescribed fire type, small live stem densities and associated carbon stocks were greater in autumn than spring, and in Low than High frequency treatments, and carbon stocks in large dead stems were greater in High than Low frequency treatments. This suggested that c. 10-yearly fires in autumn provided the most scope for maintaining future capacity to fix carbon. Nonetheless, decreases in total standing tree-based carbon stocks were not significantly different among prescribed fire treatments, suggesting tree-based carbon stocks were more influenced by prescribed fire per se than by fire season or frequency. Ó 2013 Elsevier B.V. All rights reserved.

1. Introduction Prescribed fire, the planned or ‘deliberate introduction of fire under specified fuel and weather conditions’ (Burrows et al., 2010), has been regularly used in forest management throughout Australia and elsewhere to maintain or restore species and habitat, to enhance post-logging recovery, and to reduce fuel loads and associated wildfire hazards (Fernandes and Botelho, 2003; Carter and Darwin Foster, 2004; Burrows et al., 2010; Penman et al.,

⇑ Corresponding author. Tel.: +61 3 5321 4300; fax: +61 3 5321 4166. E-mail addresses: [email protected] (L.T. Bennett), [email protected] (C. Aponte), [email protected] (K.G. Tolhurst), [email protected] (M. Löw), [email protected] (T.G. Baker). 0378-1127/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.foreco.2013.06.036

2011). In particular, the potential for prescribed fire to reduce risks from (unplanned) wildfire is highly topical given ongoing experiences of large damaging wildfires (Adams, 2013; San-MiguelAyanz et al., 2013), and given recent predictions of more frequent, extensive and severe wildfires under climate change both in temperate Australia (Bradstock, 2010; Clarke et al., 2011; King et al., 2013), and globally (Flannigan et al., 2013). In south-eastern Australia, this has led to implementation of recommendations for an expanded prescribed fire program (Parliament of Victoria, 2010; DSE, 2012), potentially involving hundreds of thousands of hectares of State land each year, and comparable with historical peaks in annual burnt area (Attiwill and Adams, 2008). Since forest fires emit greenhouse gases, regular burning of large areas of forest by prescribed fire will likely have a carbon cost

244

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255

(North and Hurteau, 2011). However, this cost might be acceptably minor if the ‘outlay’ of carbon emissions from prescribed fires was offset by a ‘return on investment’ of decreased overall emissions from wildfires (Bradstock and Williams, 2009). In North America and Europe, various studies have indicated considerable potential for carbon emission mitigation using prescribed fires (Narayan et al., 2007; Hurteau and North, 2009; Vilén and Fernandes, 2011; Ghimire et al., 2012), while others have found this potential to be limited (Campbell et al., 2011). Similarly, in temperate Australia, while some see considerable opportunities for prescribed fire to mitigate wildfire impacts on forest carbon (Adams, 2013), others have found minimal opportunity (Bradstock et al., 2012). Previous assessments of the relative carbon costs of prescribed fire regimes in temperate Australia have focused on the direct effects of burning (area burned, fuels consumed; e.g. Bradstock et al., 2012). However, research elsewhere suggests that the immediate direct effects of fire account for only a portion of actual forest carbon losses (Hurteau and North, 2009). In particular, indirect effects on stand structure and growth through scorch and death of trees and understorey plants (Busse et al., 2000; Peterson and Reich, 2001; van Mantgem et al., 2011), and associated decomposition of fire-killed biomass (Ghimire et al., 2012), can result in a sustained ‘carbon uptake legacy’ that varies with fire severity (Ghimire et al., 2012). Thus, understanding the longer-term legacies of prescribed fire on tree growth and tree-based carbons stocks is key to assessing the full carbon ‘outlays’ of different fire regimes. Indeed, Campbell et al. (2011) argue that ‘only when [fuel-reduction] treatments change the equilibrium between growth and mortality can they alter long-term carbon storage’. Data to underpin long-term assessments of prescribed fire effects on forest carbon are currently lacking for temperate Australia. Few studies have examined effects of prescribed fire on forest carbon stocks in these forests, and even fewer have considered impacts of multiple prescribed fires, or after fires of contrasting season and/or frequency. In addition, research of prescribed fire effects on tree mortality and growth in fire-tolerant forests of Australia has thus far been limited to small treatment plots of minimal replication in sub-tropical (Guinto et al., 1999) or in south-western Australia (Burrows et al., 2010). This paper presents findings from one of the most detailed and long-term studies of prescribed fire in Australia (Adams and Atti-

will, 2011). We use field measurements to examine effects of prescribed fire regimes on (standing) tree-based carbon stocks, and on long-term tree growth and mortality, in an extensive forest type of south-eastern Australia dominated by fire-tolerant eucalypts. The study has a number of unique attributes, including: (1) four prescribed fire treatments as a factorial combination of two fire seasons and two fire frequencies; (2) assessment of carbon stocks after 26 years of known prescribed fire treatment, encompassing 2–7 repeat fires; and (3) repeated measures of individual tree growth over 27 years. Our focus is carbon in standing trees because this is the predominate biomass carbon pool in Australia’s temperate eucalypt forests (Norris et al., 2010; Volkova and Weston, 2013), and is likely to be the largest carbon pool impacted by management practices like prescribed fire (Moroni, 2012). Estimates of carbon pools in forests comparable to ours have thus far been based on very few samples, with stocks in live standing trees expected to be in the range 120–240 Mg ha1 (Grierson et al., 1992; Volkova and Weston, 2013). We anticipate that carbon stocks in the soil (a focus of our ongoing research) will be appreciable but less than those in aboveground components (c. 80 Mg ha1 to 30 cm depth; Volkova and Weston, 2013), and that stocks in the understorey vegetation of our study sites will be negligible given a non-existent shrub layer. Similarly, based on Volkova and Weston (2013), we anticipate that stocks in dead standing trees (c. 9 Mg ha1) and in fallen timber (‘coarse woody debris’; c. 15 Mg ha1) will be relatively minor, and each less than 10% of stocks in live standing trees. The paper’s primary aim was to improve the empirical knowledge base for assessing and predicting effects of prescribed fire regimes on standing tree-based carbon stocks. The study’s null hypotheses were: (1) no effect of prescribed fire treatments on tree mortality or growth; (2) no effect of prescribed fire treatments on standing tree-based carbon stocks; and (3) no effect of prescribed fire season or frequency on standing tree-based carbon stocks. 2. Materials and methods 2.1. Study areas The study included five areas (known locally as the ‘Fire Effects Study Areas’, FESA) within a 25 km radius in the Wombat State For-

Table 1 Summary of the environment, stand characteristics, and fire history of the five study areas in central Victoria, Australia. Area

Latitude/Longitude Elevation (m, above sea level) a Slope (°) a Aspect (°) a Mean annual rainfall (mm) b Mean monthly max temp. (°) b Mean monthly min temp. (°) b Tree mean basal area (m2/ha) c Tree mean height (m) c Last thinning d Last wildfire d Total experimental area (ha) d Mean fire interval (yrs): AH e Mean fire interval (yrs): AL e Mean fire interval (yrs): SH e Mean fire interval (yrs): SL e a

Blakeville

Barkstead

Musk Creek

Burnt Bridge

Kangaroo Creek

37°310 S, 144°100 E 590–665 1–13 130–295 871 9–24 2–10 43 26 1964 1935 81 3.0 (6, 1987–2007) 9.5 (3, 1987–2008) 3.6 (6, 1985–2008) 9.0 (3, 1985–2005)

37°290 S, 144°050 E 635–650 0–4 120–315 901 8–23 2–10 31 28 1979 1931 19 4.0 (5, 1987–2007) 9.0 (3, 1987–2007) 3.6 (6, 1985–2005) 9.0 (3, 1985–2005)

37°280 S, 144°100 E 620–720 1–16 40–310 856 10–24 3–11 29 25 1974 1974 78 4.0 (6, 1987–2008) 16.0 (2, 1987–2004) 2.7 (6, 1986–2005) 8.5 (3, 1986–2005)

37°250 S, 144°200 E 710–760 0–15 30–270 896 7–22 2–11 43 26 1977 1953 62 5.7 (4, 1987–2007) 16.0 (2, 1987–2004) 2.7 (7, 1986–2008) 8.5 (3, 1986–2005)

37°190 S, 144°180 E 615–645 0–21 0–340 814 8–24 3–12 42 23 1975 1944 128 3.4 (6, 1987–2009) 9.0 (3, 1987–2007) 2.8 (7, 1985–2008) 9.0 (3, 1985–2005)

Range from this study’s measurement plots. From automated weather station within 4 km of each area (1986–1999 for Barkstead, 1986–2002 plus 2007–2010 for all others). c Based on measures of large stems (diameter over-bark P 20 cm at 1.3 m height) in this study’s three plots per control treatment; basal area is under-bark. d Tolhurst and Flinn, 1992. e Mean interval in years between successive prescribed fires during the experimental period (values in brackets indicate the number of prescribed fires, and the years of first and last prescribed fires); treatment abbreviations: ‘AH’ autumn High frequency, ‘AL’ autumn Low frequency, ‘SH’ spring High frequency, ‘SL’ spring Low frequency. b

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255

est, about 100 km north-west of Melbourne, Victoria, south-eastern Australia (Table 1). The areas are located on the northern (Kangaroo Creek only) and southern sides of Victoria’s Great Dividing Range, on an underlying geology of Ordovician sedimentary rock, and at elevations ranging from 590 to 760 m above sea level (Table 1). Topography of the study areas varies from mostly flat (slopes 0–4°, Barkstead) to hills of low to moderate relief (slopes up to 21°, Kangaroo Creek; Table 1), including a range of morphological types from flats and depressions to ridges and crests (morphological types of Speight, 2009). Soils are ‘stony earths’ (Blakeville only), and ‘friable earths, mottled duplex soils’ (Rowan et al., 2000), which are, respectively, Kandosols and Dermosols in the Australian Soil Classification (ASRIS, 2011). The climate is temperate, with annual rainfall in the range 814–901 mm (Table 1), the majority falling in winter and spring. Frosts are common, particularly in winter (mean of 55 days per year; Tolhurst and Flinn, 1992), and mean monthly minimum temperatures range from 2 °C (July/August) to 12 °C (January/February; Table 1). Daily summer temperatures often exceed 35 °C, with mean monthly maximum temperatures in the range 7 °C (July) to 24 °C (January/ February; Table 1). Native vegetation of the study areas is open to tall-open forest (tree heights 10 to >30 m, projective foliage cover 30–70%; Specht, 1981), with tree under-bark basal areas typically 30–40 m2 ha1 (Table 1). Open to tall-open forests are predicted to contain the majority of Victoria’s forest carbon stocks (due to their moderate productivity and extensive distribution; Kaye, 2009), and are likely to be regularly burnt under ongoing commitments to extensive use of prescribed fire on Victoria’s public land (DSE, 2012). The forests of the study areas are dominated by three co-occurring eucalypt species of different bark type: Eucalyptus obliqua L’Hér. (deep fibrous ‘stringybark’), Eucalyptus radiata Sieber ex DC. (short fibrous ‘peppermint’ bark), and Eucalyptus rubida H. Deane and Maiden (smooth ‘gum’ bark). The understorey is characterised by a sparse to non-existent shrub layer to 2–4 m height (e.g. Acacia spp.), and a ground layer dominated by Austral bracken (Pteridium esculentum (G. Forst.) Cockayne), native perennial grasses (e.g. Tetrarrhena juncea R. Br., Poa sieberiana Vickery), forbs (e.g. Gonocarpus tetragynus Labill., Viola hederacea Labill.), and rushes (Lomandra spp.; Tolhurst, 2003). The forest areas are un-evenaged, and the oldest trees are 110–120 years old (Tolhurst and Flinn, 1992). All areas were occasionally thinned to remove trees of low commercial value from the 1930s to 1960s/70s (Table 1). Detailed fire histories were not available although these forest types are prone to regular burning by wildfire (Tolhurst and Flinn, 1992), and the last known fires in the study areas prior to this experiment’s establishment were between 1935 (Blakeville) and 1974 (Musk Creek; Table 1). 2.2. Experimental design and prescribed fire treatments The study design was established in 1985, and used a randomised block design involving a long-unburnt control and four prescribed fire treatments randomly allocated within each of the five study areas (total of 25 treatment areas). The available experimental areas ranged from 19 ha (Barkstead) to 128 ha (Kangaroo Creek; Table 1), so individual treatments ranged from 3 to 35 ha in area (Tolhurst and Flinn, 1992). The four prescribed fire treatments involved a factorial combination of two fire seasons (autumn or spring), and two fire frequencies (nominally every 3 or 10 years); that is, autumn High frequency (‘AH’), autumn Low frequency (‘AL’), spring High frequency (‘SH’), and spring Low frequency (‘SL’). Nominal prescribed fire intervals of three and ten years were chosen to represent, respectively, the shortest interval for sufficient recovery of surface fuels to carry a fire in these forests, and the likely return interval of prescribed fire based on local fire management practice. Due to

245

seasonal and operational constraints, mean prescribed fire intervals ranged from 2.7 to 5.7 years in the High frequency treatments, and 8.5–16 years in the Low frequency treatments (Table 1). The date of last prescribed fire ranged from March 2004 to March 2009 (that is, about 7.5 to 2.5 years prior to this study’s carbonstock measurements in late 2011; Table 1). Prescribed fires in all treatments in this study were considered mild (Tolhurst and Flinn, 1992) as indicated by Forest Fire Danger Index (FFDI) means of 6 7 (low to moderate fire danger rating; Luke and McArthur, 1978). Thus, fire intensities were generally less than 500 kW m1, and involved little overall canopy scorch (DSE, 2003). Field observations of individual prescribed fires indicated flame heights were in the range 0.1 to 1.3 m, and did not differ between fire treatments (Table 2). However, tree scorch heights were often >10 m, and greater mean and maximum scorch heights indicated greater overall severity of Low than High frequency fire treatments (no fire season effects; Table 2). Mean area burnt (79–92%), and mean surface fuel consumed (40–53%) did not significantly differ between prescribed fire treatments (Table 2). Nonetheless, both the mean Forest Fire Danger Index (FFDI; McArthur, 1967) and the mean Soil Dryness Index (SDI; Mount, 1972) indicated significantly drier fire conditions for autumn versus spring fires (Table 2).

2.3. Plot-based tree measures Trees and stumps were measured (August to November 2011) within three circular plots (18 m radius, c. 0.1 ha) per treatment area (total of 75 plots). ‘Trees’ were defined as upright or leaning, live or dead woody stems that were >1.3 m height, and were rooted in the ground (to distinguish from fallen trees that were classified as coarse woody debris). ‘Stumps’ were 6 1.3 m height, and were the in situ lowest part of a cut tree stem (from prior logging), or of a naturally broken stem. Large trees were defined as P 20 cm diameter (over-bark) at breast height (‘dbh’, 1.3 m height), and large stumps as P 20 cm diameter at 0.3 m height. Stumps were assumed dead, although any coppice/epicormic regrowth stems from stumps were assessed as individual live/dead trees, as were multiple stems of the same tree. All large live trees in the 0.1 ha plots were assessed for: species, dbh over-bark (mm), and bark thickness at breast height. Bark thickness was measured using a ‘Gill-type’ needle gauge (Gill et al., 1982) as the mean radial depth (mm, surface to cambium) of four points around the stem (north, south, east, west sides), taking care to avoid bark furrows. Species of large dead trees were not identifiable, and large dead trees were only assessed for dbh (overbark or under-bark as relevant), and for remaining cross-sectional area at dbh (%). Large stumps, assumed to be cylindrical, were measured for diameter (mm) at 0.3 m height (i.e. assuming the full cross-sectional area was present) and total height (m), and assessed for the proportional volume of the full cylinder remaining (to the nearest 10%). Small trees (live and dead; <20 cm dbh) and stumps (<20 cm diameter at 0.3 m height) were counted in two size classes (<10 cm, 10 to < 20 cm diameter) in the north-east and south-west quadrants of each plot (total 0.05 ha). In addition, ‘seedlings’, defined as small live trees < 1 cm in mean diameter, and usually < 2 m in height, were counted in each of these quadrants. 2.4. Estimation of tree-based biomass and carbon stocks Biomass of large tree components (under-bark stem, bark, branch, foliage) was estimated using extant additive biomass equations for open-forest eucalypts (Bi et al., 2004). The equations were of the form:

246

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255

Table 2 Characteristics of prescribed fire treatments (autumn High frequency ‘AH’, autumn Low frequency ‘AL’, spring High frequency ‘SH’, spring Low frequency ‘SL’). Values are the mean (±SE) of ‘n’ observed prescribed fires (in square brackets) across the five study areas. Treatment AH Flame height (m) Mean Maximum Minimum

Significant effects AL

SH

a

SL

b

Scorch height (m) c Mean Maximum Minimum Area burnt (%) d Surface fuel consumed (%) Forest fire danger index f Soil dryness index g

0.34 ± 0.04 [20] 0.98 ± 0.23 [17] 0.11 ± 0.01 [13]

e

5.72 ± 1.04 [15] 10.71 ± 1.72 [15] 2.83 ± 0.69 [11] 87 ± 4 [17] 40 ± 5 [14] 6 ± 1 [27] 111 ± 7 [27]

0.45 ± 0.07 [10] 1.29 ± 0.39 [10] 0.14 ± 0.03 [8] 7.49 ± 1.34 [9] 14.56 ± 2.70 [9] 2.69 ± 0.66 [7] 92 ± 2 [10] 41 ± 7 [8] 7 ± 2 [12] 125 ± 12 [12]

0.43 ± 0.06 [33] 0.79 ± 0.09 [29] 0.21 ± 0.05 [22]

0.41 ± 0.05 [15] 0.97 ± 0.10 [15] 0.13 ± 0.02 [10]

4.42 ± 0.61 [29] 7.76 ± 1.11 [28] 2.71 ± 0.48 [20] 79 ± 5 [29] 53 ± 3 [19] 4 ± 1 [34] 30 ± 2 [34]

6.79 ± 1.55 [15] 12.78 ± 1.86 [15] 4.17 ± 1.29 [10] 87 ± 3 [15] 53 ± 5 [10] 3 ± 0 [15] 19 ± 2 [15]

— — — Fq* Fq** — — — S*** S***, SxFq*

a Significant effects of prescribed fire season (‘S’) and frequency (‘Fq’) and their interaction as indicated by a Linear Mixed Model (*P < 0.05; **P < 0.01; ***P < 0.001; — no significant effects detected). b Visual estimation of average flame height at each of 16 posts either in a line at 2 m intervals, or on a 5  10 m grid (Tolhurst and Flinn, 1992). c Visual estimation of scorch height at a minimum of 20 systematically located points on a 50  100 m grid (or smaller as dictated by the treatment area); measured at each point as the highest tree scorch (as measured using distance and angle) within a 10 m radius (Tolhurst and Flinn, 1992). d Proportion area burnt at a minimum of 20 systematically located points on a 50  100 m grid (or smaller as dictated by the treatment area); measured at each point (to the nearest 0.10 m) as the proportion of the length of a 10 m transect that intersected with burnt patches (Tolhurst and Flinn, 1992). e Proportional decrease in oven-dry mass (105 °C) of surface fuel (<25 mm in minimum cross-sectional area) measured in 15 randomly located 0.1 m2 quadrats before and after prescribed fire (Tolhurst and Flinn, 1992). f McArthur Forest Fire Danger Index (Luke and McArthur, 1978), based on weather data in the open (air temperature, relative humidity, rainfall, wind speed) from automated weather stations within 4 km of each area, and, within each study area, using a Bacharach sling psychrometer (air temperature, relative humidity). g Soil Dryness Index (Mount, 1972) based on weather near and within each study area as above.

Y i ¼ ebi0 Dbi1 þ i ; where Yi is the biomass of component i, D is over-bark diameter at breast height (‘dbhob’), bi0 and bi1 are defined coefficients, and ei is the component error term (Bi et al., 2004). Species-specific equations were available for both E. obliqua and E. radiata, but it was necessary to substitute Eucalyptus dalrympleana equations for E. rubida, and Acacia dealbata equations for (occasional) Acacia melanoxylon R. Br. (Bi et al., 2004). Significant decreases in the bark thickness of E. obliqua with burning (Section 3.1), indicated probable underestimation of predicted biomass components of burnt trees from equations based on ‘dbhob’ (i.e. that smaller over-bark diameters in prescribed fire than control trees would erroneously predict less biomass). Thus, for live large E. obliqua in prescribed fire treatments only, biomass in stems (under-bark), foliage and branches was corrected for bark loss by using estimated ‘non-burnt’ dbhob (i.e. the dbhob had the trees not been burnt), as follows: dbhub (dbh under-bark) was calculated for all E. obliqua as measured dbhob minus measured bark thickness; a strong linear relationship was established between dbhob and dbhub of E. obliqua in control treatments across all areas (dbhob = 57.09 + 1.05 * dbhub; R2 0.99; n = 141); estimated ‘nonburnt’ dbhob of burnt trees was calculated using the dbhob to dbhub relationship, and then used in the above-mentioned component biomass equations (Bi et al., 2004). Bark biomass loss due to burning was estimated in two ways: (i) the cylindrical volume to breast height (1.3 m) of the estimated ‘non-burnt’ tree (i.e. using the estimated ‘non-burnt’ dbhob) minus the cylindrical volume to 1.3 m of the burnt tree (i.e. using the measured dbhob), multiplied by an average bark density for E. obliqua >20 cm dbhob (212 kg/m3; Attiwill, 1979); and (ii) predicted ‘non-burnt’ bark biomass using the ‘non-burnt’ dbhob in the above-mentioned bark allometric equation minus predicted bark biomass of the burnt tree using the measured dbhob in the same equation. Since the cylindrical method gave the lower estimates of the two (see Section 3.1), it was used to deduce the remaining bark biomass of the burnt trees; that is, by subtracting bark loss

from the predicted bark biomass using the estimated non-burnt dbhob in the bark equation. Under-bark stem biomass of dead large trees was calculated using species-specific allometric equations (as above), with tree species assigned based on the most common live species in that plot. Dead stem biomass estimates were then adjusted for remaining cross-sectional area, and for minor decay using a multiplier of 0.85. This multiplier was based on an average ratio of sound-dead to fresh-wood densities of over 50 pieces of coarse woody debris that we sampled across the study areas, and was consistent with similar ratios in other temperate forests (e.g. Coomes et al., 2002; Grove et al., 2009). Similarly, biomass of large stumps was calculated as the volume of the assumed cylinder, corrected for remaining volume, and multiplied by an average density of sound-dead wood (459 kg m3, based on the above-mentioned coarse woody debris samples). Small trees were counted but not measured, so an estimated biomass of a single stem in each of the small size classes (<10 cm, 10 to < 20 cm dbhob) was used in total biomass calculations. The 10 to <20 cm dbhob size class was within the range of available allometric equations, so biomass of live trees in this class was estimated as follows: above-mentioned additive biomass equations (Bi et al., 2004) for E. obliqua and E. radiata (the most frequent tree species) were used to predict component biomass (under-bark stem, bark, branch, foliage) for each 1 cm-increment from 10 to 20 cm dbhob; biomass of all components was summed to give total tree biomass by diameter increment within species; and these values were averaged to give an estimated biomass of 108 kg per tree. Biomass of dead trees in this size class was calculated in a similar way, although only the under-bark stem component was included, and a decay multiplier of 0.85 was applied (as above), to give an estimated biomass of 56 kg per tree. The biomass of a live tree in the smallest size class (<10 cm dbhob) was estimated using a generic equation for small eucalypts from Keith et al. (2000), of the form:

LnðYÞ ¼ 1:07 þ ð2:89LnðHÞÞ

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255

where Y is the biomass of a single tree, and H is the tree height (here, based on field observations, estimated as 2 m). This gave an estimate of 2.5 kg per tree, which became 2.1 kg for dead trees using a 0.85 decay multiplier. As for large stumps, biomass of representative small stumps was estimated by multiplying a cylindrical volume (assuming 20 cm height for the <10 cm diameter class, and 60 cm height for the 10 to <20 cm class, and using the midrange diameters) by an average density of sound-dead wood (459 kg m3, as above). This gave estimates of 0.2 kg per stump in the <10 cm diameter class, and 4.8 kg per stump in the 10 to <20 cm diameter class. Tree (live, dead) and stump biomass by components and size classes were summed for each plot (i.e. using individual large tree/stump values, and multiplying single small tree/stump estimates by counts), and then multiplied by 0.50 to convert mass to carbon content. These data were then converted to carbon (C) in Mg ha1 by dividing by plot area, which was corrected for slope using the slope-adjusted distances of two perpendicular transects that ran through the plot centre. 2.5. Repeated measurements of individual trees across treatments In addition to the above plot-based measurements, 775 individual trees in selected treatments were repeatedly assessed from 1985 to 2012 for stem diameter and bark thickness. These trees, mostly E. obliqua, were selected in 1985 for good form (straight stem, healthy crown), and to cover six size classes: 0–10, 10.1– 20, 20.1–30 cm, 30.1–50, 50.1–70, and >70 cm dbhob (Chatto et al., 2003). That is, trees were selected to represent a range of dominance classes rather than a random sample. Not all area-treatment combinations could be sampled this intensively; instead, trees were selected in all treatments at Blakeville, in control, AH, and SL treatments at Barkstead and Kangaroo Creek, and in control, SH, and AL treatments at Musk Creek and Burnt Bridge. This sampling design ensured that control versus prescribed fire comparisons were based on replicated observations of a range of prescribed fire conditions. In addition, to allow for some evaluation of fire effects on other bark types, at least five E. rubida trees in each of the six size classes per five treatments were selected for repeated measurement at Blakeville only. Prior to the first prescribed fires in 1985, the selected trees were assessed for competition using the zone count method (10 m2 ha1 optical wedge; Opie, 1968), and were measured for dbhob and bark thickness at breast height after ‘some initial trimming to remove loose bark’ (Chatto et al., 2003). These measures were repeated after the first prescribed fire, on a regular basis to 1999, and in 2012. Pre-fire 1985 bark thickness measures were based on two measures at right angles using a Swedish bark gauge, but, thereafter, on four measures around the tree using a Gill-type needle gauge (Chatto et al., 2003). The initial bark trimming at breast height was not consistent with tree measurements in the plots (Section 2.3), and could lead to underestimation of prescribed fire effects on bark thickness of individual trees. Thus, fire effects on individual tree growth were only examined using under-bark dbh (i.e. dbhob minus measured bark thickness). In addition, overall effects of prescribed fire on bark thickness of individual trees were only examined using bark thickness at 50 cm height (i.e. no trimming), which was assessed in 2012. 2.6. Statistical analyses Effects of prescribed fire treatments on plot-based tree and stump measures (e.g. tree count ha1, carbon stock Mg ha1) were tested using the General Analysis of Variance models of GenStat (14th edition, VSN International Ltd, Hemel Hempstead, UK), with area as a random factor, treatment as a fixed factor, and plot nested

247

within treatment by area. In addition, a ‘factorial plus added control’ model (i.e. ‘season frequency’ nested within prescribed fire; Payne, 2011) was used to assess overall effects of prescribed fire treatment (i.e. control versus prescribed fire), of fire season (autumn versus spring), and of fire frequency (High versus Low). REML Linear Mixed Models (GenStat 14th edition) of a similar structure were used if values were missing (e.g. no trees of a particular size class), or if the analysis involved plot-level covariates. Assumptions of normality and variance homogeneity were checked, and dependent variables transformed as necessary (ln or (Y + 0.5)^0.25 for counts, arcsine for proportions; Quinn and Keough, 2002). In addition, further checks of fire effects were made using logit transformations of proportional data (Warton and Hui, 2011), and using GenStat’s Permutation Test (4999 random permutations). To account for potential pre-treatment differences in stand stocking between treatment plots, all analyses of fire treatment effects on carbon stocks were re-run using plot-level live and dead large stem density as a covariate. Probability of death by 2012 of the 775 selected individual trees was assessed using GenStat logistic regression models (logit transformation) with size class, competition, study area, and prescribed fire treatment as predictor variables. Here, an overwhelming effect of size class resulted from high mortality of trees 6 20 cm; as such, the 0–10 and 10.1–20 cm size classes were excluded from the analyses below of individual tree growth. Repeated measures of individual live trees (>20 cm dbhub) produced predominantly linear relationships between dbhub and time (1985–2012). Thus, overall effects of prescribed fire treatments on individual tree growth were assessed using slopes of the dbhub/ time relationships as a dependent variable in the above-mentioned REML Linear Mixed Models (GenStat 14th edition), with initial dbhub and initial zone-count competition as covariates, and/or size class as a fixed factor. Prior to these analyses, and acknowledging the potential for measurement error associated with multiple assessors over the study’s 27 years, individual dbhub versus time relationships were examined (blind to treatment), and points or whole relationships were excluded using the following highly conservative criteria: (1) exclude a single point if it was a clear outlier in the tree’s overall dbhub/time relationship, but only if dbhub decreases/increases between that point and the preceding and following points were considered improbable on the basis that they were more than four times the mean diameter increment for the corresponding time intervals as indicated by all relationships for that species and size class (after Murphy et al., 2010); (2) exclude a whole relationship if there were fewer than 8 points remaining after exclusions based on the first criteria; and (3) exclude a whole relationship if the 2012 measurement point was excluded due to an improbable decrease relative to the preceding point (i.e. include only those trees that had credible growth for the entire measurement period). Application of these criteria led to the exclusion of 21 live E. obliqua >20 cm dbhub (335 remaining), and 5 live E. rubida >20 cm dbhub (76 remaining).

3. Results 3.1. Tree bark thickness and bark carbon loss Burning significantly decreased mean bark thickness at breast height of large live E. obliqua in the plots (39 mm ± 1.7 SE in controls, 31 mm ± 0.7 across all fire treatments, P < 0.001), but had no effect on the mean bark thickness of either E. radiata (controls: 22 mm ± 1.2, fire treatments: 21 mm ± 0.4), or E. rubida (controls: 23 mm ± 1.0 controls, fire treatments: 21 mm ± 0.5; Supplementary data, Fig. S1a). In addition, decreases in relative bark thickness (bark thickness/dbhub) were greater in the High than Low fre-

248

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255

240 Large live stems Large live bark, branch, foliage Small live stems: A>S*, L>H*

C (Mg ha-1)

200

a

160 120 80 40 0 Control

AH

AL

SH

SL

20

C (Mg ha-1)

b

Large dead stems: Cont>F**, H>L* Small dead stems Large and small stumps: L>H*

16 12 8 4 0 Control 320

AL

SH

SL

Total tree live and dead: Cont>F* Total tree live Total tree dead: Cont>F**

280 240

C (Mg ha-1)

AH

c

200 160 120 80 40 0 Control

AH

AL

SH

SL

Treatment Fig. 1. Estimated tree carbon (Mg ha1) in plots, by large (P20 cm dbhob) and small (<20 cm dbhob) live stems, including non-stem components of large live trees (a), dead stems and stumps (b), and all live and dead tree components (c, where ‘total tree live’ is the sum of components in ‘a’; ‘total tree dead’ is the sum of components in ‘b’; and ‘total tree live and dead’ is the sum of these two). Treatments were a non-burnt control, prescribed fires in autumn at High frequency (‘AH’), in autumn at Low frequency (‘AL’), in spring at High frequency (‘SH’), and in spring at Low frequency (‘SL’). Values are the means of 15 plots with 95% confidence intervals. Asterisks indicate overall effects (*P < 0.05, **P < 0.01) of prescribed fire (control versus prescribed fire, ‘F’), of fire season (‘S’ versus ‘A’), and of fire frequency (‘H’ versus ‘L’).

quency fire treatments for both E. obliqua (0.15 versus 0.17, P < 0.05) and E. rubida (0.10 versus 0.12, P < 0.05; Supplementary data, Fig. S1b). Consistent with the plot data, bark thickness at 50 cm height of selected individual E. obliqua was significantly decreased by burning (32 mm ± 0.5 SE across all fire treatments, 40 mm ± 1.4 in controls, P < 0.05), and relative bark thickness was also significantly less in High (0.065) than Low frequency fire treatments (0.082, P < 0.05; data not shown). Estimates of mean losses of bark carbon mass from E. obliqua in fire treatment plots ranged from 0.21 Mg ha1 (± 0.02 SE, cylindrical method) to 0.39 Mg ha1 (± 0.03 SE, allometric-based method). These relatively low values were not significantly different among fire treatments, and indicated that more accurate estimation of these losses was not warranted. 3.2. Live tree-based carbon stocks Total live tree-based carbon stocks in plots ranged from 69 to 318 Mg ha1 (overall mean 174 Mg ha1). Against this variable

background, we did not detect a significant difference in live tree-based carbon between control (191 Mg ha1 ± 13 SE) and prescribed fire treatments (170 Mg ha1 ± 4; P 0.087; Fig. 1c, Table 3). Similarly, statistical models did not detect an overall effect of prescribed fire on carbon in the main component of this pool, namely, live large stems (control 138 Mg ha1; fire treatments 120 Mg ha1; P 0.057; Fig. 1a, Table 3); although we note that the 95% confidence interval for the difference between control and prescribed fire means did not include zero (0.3–35.6 Mg ha1; Table 3). No significant difference in total live tree-based carbon stocks between control and prescribed fire treatments reflected no overall significant effects of prescribed fire on plot-level large and small live stem densities (Fig. 2). While effects of prescribed fire per se on live tree-based carbon in plots were muted, effects of the type of prescribed fire were more clearly significant. Mean carbon stocks in small live stems were significantly greater in autumn (9.0 Mg ha1) than spring fire treatments (6.4 Mg ha1, P 0.041; Fig. 1a; Table 3), and in Low (9.3 Mg ha1) than High frequency treatments (6.1 Mg ha1, P 0.013; Fig. 1a; Table 3). This reflected greater live stem densities

249

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255 Table 3 Summary of the significance of the main effects of prescribed fire treatments on carbon stocks in standing tree-based carbon pools. a

Pool

Effect

Total tree live and dead Total tree live Total tree dead Large live stems Large dead stems

Cont > Fire Cont = Fire Cont > Fire Cont = Fire Cont > Fire High > Low Aut > Spring Low > High Low > High

Small live stems Large and small stumps

Difference (Mg C ha1)

b

P-value

c

Mean

95% CI

Without covariate

With covariate

25.0 21.7 3.3 18.0 3.5 2.4 2.6 3.2 1.0

2.1–47.8 -1.1–44.4 1.0–5.7 0.3–35.6 1.5–5.4 0.6–4.1 0.3–4.9 0.9–5.6 0.2–1.8

0.049 0.087 0.011 0.057 0.004 0.021 0.041 0.013 0.025

0.042 0.073 0.009 0.059 0.003 0.020 0.034 0.010 0.021

a

Effect from ‘Factorial plus added control’ ANOVA model without covariate. Potential effects were prescribed fire (Control versus prescribed Fire), fire season (Autumn versus Spring), and fire frequency (High versus Low). b First listed treatment in ‘Effect’ column minus second listed (e.g. Control minus Fire treatments). c ‘Without covariate’ indicates the significance of the model effect in the ‘Effect’ column; ‘With covariate’ indicates the significance of the effect using the same model but with the plot-level number of live and dead large stems as a covariate.

in autumn than spring and in Low than High frequency treatments (Fig. 2a), combined with significantly less small-stem mortalities in these treatments (Fig. 3a).

3.3. Dead tree-based carbon stocks Mean dead tree-based carbon in plots was significantly less in prescribed fire (8.7 Mg ha1 ± 0.6 SE) than control treatments (12.0 Mg ha1 ± 2.1; P 0.011; Fig. 1b, Table 3). Most of this decrease

was due to significant differences in carbon in large dead stems, which contained 55% of the standing dead carbon in control treatments, and 36% in prescribed fire treatments (Fig. 1b). In turn, of the total carbon stock in large live and dead stems, proportionally more was in dead stems in control (4.9%) than fire treatments (2.6%; P 0.048). Despite differences in dead tree-based carbon mass, there were no detectable overall effects of prescribed fire on plot-level large and small dead stem densities (Fig. 2), nor on proportional numbers of small and large dead stems (Fig. 3).

700

Small stem density (ha-1)

600

a

Live: AL>SH, A>S**, L>H* LRGN_BackT Dead: AL>SH, A>S*, L>H* DRGN_BackT

500 400 300 200 100 0 Control

AH

AL

SH

SL

400

b

Live LOWN_BackT

Large stem density (ha-1)

350

Dead: H>L*

300 250 200 150 100 50 0 Control

AH

AL

SH

SL

Treatment Fig. 2. Densities of small (a; dbhob < 20 cm) and large (b; dbhob P 20 cm) stems, both live and dead, in plots. Treatments were a non-burnt control, prescribed fires in autumn at High frequency (‘AH’), in autumn at Low frequency (‘AL’), in spring at High frequency (‘SH’), and in spring at Low frequency (‘SL’). Values are the means (fourth-root back-transformed) of 15 plots with 95% confidence intervals. Significant differences between individual fire treatments are as listed (Tukey post hoc test, P < 0.05), and asterisks indicate overall significant effects of fire season (‘S’ versus ‘A’), and of fire frequency (‘H’ versus ‘L’); *P < 0.05, **P < 0.01.

250

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255

In addition to overall effects of prescribed fire, carbon stocks in standing dead components were affected by the type of prescribed fire. Carbon in large dead stems was significantly greater in High (4.3 Mg ha1 ± 1.0 SE) than Low frequency fire treatments (1.9 Mg ha1 ± 0.5; P 0.021; Fig. 1b, Table 3), consistent with significantly greater densities of large dead stems (Fig. 2), and of proportional numbers of large dead stems (Fig. 3). Low overall numbers of dead large stems prevented formal analyses of fire effects by size class; nonetheless, the majority of all large stem deaths were of stems 20–50 cm dbhob, and there was a tendency for more deaths of stems >50 cm dbhob in the High than Low frequency fire treatments (Fig. 3b). In contrast to large dead stems, mean carbon mass in stumps was significantly greater in Low (4.0 Mg ha1 ± 0.4 SE) than High frequency fire treatments (3.0 Mg ha1 ± 0.3, P 0.025; Fig. 1b, Table 3).

3.4. Total live and dead tree-based carbon stocks When all standing dead and live tree components were combined, we detected significantly less carbon mass in prescribed fire (178 Mg ha1 ± 4 SE) than control treatments (203 Mg ha1 ± 13; P 0.049; Fig. 1c, Table 3). This equated to a mean difference of 25 Mg ha1 (c. 12% of total tree-based carbon in control treatments), although the 95% confidence interval for this difference was wide (2–48 Mg ha1; Table 3). Including plot-level live and dead large stem density as a covariate strengthened the significance of the prescribed fire effect on total tree-based carbon (P

0.042), as was generally true for tests of stocks in other tree-based pools (Table 3). Despite significant effects of prescribed fire frequency and season on various components of live and dead tree carbon mass, we did not detect any effects of fire type on the total carbon stock in all the live and dead tree components combined (Fig. 1).

3.5. Individual tree mortality and growth Mortality of small (<20 cm dbhub in 1985) individual E. obliqua between 1985 and 2012 ranged from 27% to 49%, and was not significantly affected by prescribed fire (Fig. 4a). In contrast, mortality of large individual E. obliqua was only 1% in control treatments, increasing marginally but significantly in fire treatments (3%), largely due to 7% mean mortality in the SL treatments (Fig. 4a). Mortality of the smooth-barked E. rubida at Blakeville (only) over the same period tended to be higher, particularly in the fire treatments: 46% of small stems in the control versus 56% in fire treatments, and 5% of large stems in the control versus 7% in fire treatments (Fig. 4b). With initial dbhub as a covariate, mean under-bark diameter increment from 1985 to 2012 of large individual E. obliqua was significantly less in fire treatments (4.7 mm year1) than control treatments (5.3 mm year1, P 0.028), and was not obviously affected by type of fire treatment (Fig. 5a). The significance of the overall effect of prescribed fire on E. obliqua diameter increment was increased with the inclusion of initial zone-count competition as an additional covariate (P 0.026). Decreased mean diameter

80

Dead small stems (%)

70

a

dbhub < 10 cm dbhub 10_20

SH>AL

60

S>A*, H>L*

50 40 30 20 10 0 Control

AH

AL

SH

SL

16

Dead large stems (%)

14 12

b

dbhub 20_30 cm dbhub 30_50 cm dbhub 50_70 cm dbhub > 70 cm

H>L*

10 8 6 4 2 0 Control

AH

AL

SH

SL

Treatment Fig. 3. Percentages of small (a) and large (b) stems that were dead in plots by diameter classes. Treatments were a non-burnt control, prescribed fires in autumn at High frequency (‘AH’), in autumn at Low frequency (‘AL’), in spring at High frequency (‘SH’), and in spring at Low frequency (‘SL’). Total values are the mean proportion (arcsine back-transformed) of 15 plots with 95% confidence intervals. Significant differences between individual fire treatments in total dead percentages are as listed (Tukey post hoc test, P < 0.05), and asterisks indicate overall significant effects of fire season (‘S’ versus ‘A’), and of fire frequency (‘H’ versus ‘L’) on total dead percentages (*P < 0.05).

251

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255

80 All stems Small stems Large stems: F>Cont*, L>H*

Dead stems (%)

70 60

a

50 40 30 20

7 2

1

10

2 0

0 Control 80

Dead stems (%)

70 60

AH

AL

SH

SL

All stems Small stems Large stems

b

50 40 30 20

10

7

5

10

7

6

0 Control

AH

AL

SH

SL

Treatment Fig. 4. Percentage deaths between 1985 and 2012 of selected individual E. obliqua trees (a), and E. rubida trees (b), both small (initial dbhub <20 cm) and large stems (dbhub P20 cm). Treatments were a non-burnt control, prescribed fires in autumn at High frequency (‘AH’), in autumn at Low frequency (‘AL’), in spring at High frequency (‘SH’), and in spring at Low frequency (‘SL’). For E. obliqua, values are the means of 5 (control) to 3 (all others) plots per treatment with 95% confidence intervals. For E. rubida, values are the percentage deaths in just one plot per treatment at Blakeville. For clarity, percentage deaths of large stems are provided within each figure. Asterisks indicate overall effects (*P < 0.05) of prescribed fire (control versus prescribed fire, ‘F’), of fire season (‘S’ versus ‘A’), and of fire frequency (‘H’ versus ‘L’) on percentage deaths of E. obliqua.

increment with prescribed fire was most obvious for the smallest of the large E. obliqua, but was only significant in the 30–50 cm dbhub size class (4.1 versus 5.3 mm year1, P 0.043 with one covariate, P 0.032 with two covariates; Fig. 5b). Nonetheless, the interactive effect of fire treatment and size class was not significant (P 0.094). Consistent with the effects across all large E. obliqua size classes, there were no detectable effects of fire type (season or frequency) on the mean diameter increments within each size class. Under-bark diameter increments of individual E. rubida at Blakeville only (i.e. replicate treatments not sampled) were highly variable, with treatment means ranging from 4.3 mm year1 in the control to 5.9 mm year1 in the ‘AL’ treatment (data not shown). Based on tree plot measures, 26% of E. obliqua stems were in the 30–50 cm dbhub size class, the most of any size class (and the size class with mean diameter increment significantly decreased by prescribed fire treatment, as above). ‘Growing’ all stems in this size class in the plots by the ‘fire treatment rate’ of 4.1 mm year1 (above) led to mean plot tree carbon mass (all components) in this size class of 17.9 Mg ha1 after 10 years, and 21.5 Mg ha1 after 20 years. In contrast, growing stems in this size class by the ‘control rate’ of 5.3 mm year1 led to mean plot tree carbon mass of 18.9 Mg ha1 after 10 years and 23.7 Mg ha1 after 20 years. That is, prescribed fire treatments had the potential to decrease mean tree carbon mass in this size class by about 1.0 Mg ha1 over 10 years, and 2.2 Mg ha1 (9%) over 20 years (i.e. 0.1 Mg ha1 year1). 4. Discussion 4.1. Prescribed fire effects on tree mortality Small-stem (<20 cm dbhub) mortality rates in this study (across-treatment mean of 35% of individual E. obliqua over

27 years) were high compared with other temperate eucalypt forests (12% over 20 years: Burrows et al., 2010), but, consistent with comparable studies (Burrows et al., 2010), were not obviously influenced by prescribed fire. Rather, our plot data indicated that while small-stem mortalities did not change with prescribed fire per se, they increased with more frequent fire, and with spring rather than autumn fire. Increased small-stem mortality with more frequent fire is consistent with observations of other eucalyptdominated forests (Abbott and Loneragan, 1984; Wilkinson and Jennings, 1993; Guinto et al., 1999), and is presumably related to insufficient time between fires for recovery of protective bark thickness (Gill, 1980; Wilkinson and Jennings, 1993; Lawes et al., 2013). Nonetheless, greater small-stem mortalities after spring than autumn fires were not consistent with indications of milder fire conditions in spring (as indicated by significantly lower FFDI and SDI), but were presumably associated with less rainfall in the summer months after spring fires, and thus with comparatively greater potential for negative interactive effects between fire damage and available water (Fensham et al., 2008). Mean mortality of large individual E. obliqua in controls of 1% over 27 years (0.04% yr1) was low relative to non-fire mortality rates of large fire-tolerant trees in, for example, conifer forests of south-western USA (0.5–1% yr1; van Mantgem et al., 2011), but was consistent with low background mortality rates of large eucalypts in other studies (Abbott and Loneragan, 1983; Burrows et al., 2010). Prescribed fires marginally but significantly increased mean mortality rates of large individual E. obliqua to 3% over 27 years (0.1% yr1), which again seems negligible relative to increased mortality rates with prescribed fire of large fire-tolerant pines elsewhere (4.6% yr1; van Mantgem et al., 2013). Nonetheless, other medium-term studies of large fire-tolerant eucalypts have also detected only minor impacts on mortality of single or repeated prescribed fires (Abbott and Loneragan, 1983; Burrows et al., 2010).

252

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255

Diam. growth (mm yr-1)

7

a

Control (5.3) > Fire (4.7)*

6 5 4 3 2 1

[93]

[57]

Control

AH

[65]

[62]

[58]

AL

SH

SL

0

Treatment 9

Diam. growth (mm yr-1)

8 7

b

Control Cont mm/yr Fire mm/yr Burn

6

*

5 4 3 2 1

[18] [48]

[30] [78]

[38] [89]

20 to 30

30 to 50

50 to 70

[7]

[27]

0 70+

Size Class (initial dbhub; cm) Fig. 5. Mean under-bark diameter growth rate between 1985 and 2012 of large (initial dbhub P20 cm) individual E. obliqua trees, by prescribed fire treatment (a) and by size class within control treatments and across all fire treatments (b). Treatments were a non-burnt control, prescribed fires in autumn at High frequency (‘AH’), in autumn at Low frequency (‘AL’), in spring at High frequency (‘SH’), and in spring at Low frequency (‘SL’). Values are the means with 95% confidence intervals of diameter increments of n individuals (indicated in square brackets). Asterisks indicate an overall significant effect of fire on diameter growth (a, *P 0.028), and a significant difference between control and prescribed fire treatment trees in the 30–50 cm size class (b, *P 0.043). All statistical analyses included initial dbhub as a covariate (inclusion of initial zone-count competition as an additional covariate increased the significance of effects).

Despite low overall mortality rates, both our plot and individual tree data indicated significant effects of prescribed fire frequency on large tree mortality. Percentages of dead large stems in plots were significantly greater in High versus Low frequency treatments. This is consistent with (very few) other observations of increased large eucalypt mortality with increasing fire frequency (Collins et al., 2012). Our study indicates this could be associated with significant decreases in relative bark thickness of both E. obliqua and E. rubida in High versus Low frequency treatments, which would increase potential (particularly for the thinner-barked E. rubida) for the cambium to be damaged by High fire temperatures (Gill and Ashton, 1968; Abbott and Loneragan, 1983). In apparent contrast to the plot data, mortality of large individual E. obliqua was significantly greater in Low than High frequency treatments (largely due to greater mortality in the SL treatment). This could be an effect of greater mean fire severity, as indicated by greater scorch height, in the Low frequency treatment (van Mantgem et al., 2013). Lack of agreement between plot- and tree-level data on frequency effects could be attributed to greater proportional contributions of eucalypts other than E. obliqua to total dead counts in the plots. In any case, our findings illustrate the need to balance the likely effects of frequency with severity on large tree mortality in prescribed fire regimes. Interpreting the significance of marginally increased large tree mortality after low-intensity prescribed fires (either more or less frequent) requires consideration of the likely mortalities following

infrequent but higher-intensity wildfire (van Mantgem et al., 2013). Based on surprisingly few available observations, overall mortalities in comparable fire-tolerant eucalypt forests after wildfire have been in the range 5–10% (Abbott and Loneragan, 1983; Strasser et al., 1996; Benyon and Lane, 2013), but increased to between 25% and 50% following the most severe wildfire (70–100% of overstorey crowns burnt; Benyon and Lane, 2013). As such, there seems little scope for prescribed fire regimes to markedly influence the mortality rates of fire-tolerant eucalypts, unless they markedly mitigate, or indeed exacerbate (Strasser et al., 1996), the damaging effects of severe wildfires. Such interactions between prescribed fires and wildfire on large eucalypt mortality remain un-examined. 4.2. Prescribed fire effects on tree growth We detected significantly less mean diameter growth of E. obliqua individuals in prescribed fire than control treatments over 27 years, irrespective of fire treatment. This effect is consistent with reduced diameter growth of large Corymbia maculata (syn. Eucalyptus maculata) after 6 years of annual burning in northern New South Wales (Floyd, 1966), but is seemingly at odds with other studies that found no clear effect of prescribed fire on diameter growth of fire-tolerant eucalypts (Abbott and Loneragan, 1983; Guinto et al., 1999; Burrows et al., 2010). Nonetheless, these previous studies had potentially less power to detect growth effects for a range of reasons including less replication of fire treat-

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255

ments, non-balanced designs and/or small experimental areas(Abbott and Loneragan, 1983; Guinto et al., 1999; Burrows et al., 2010). Based on broader evidence, our detected decreases in tree growth associated with prescribed fire could be due to decreased availability of nutrients related to the repeated combustion of leaf litter and humus (Abbott and Loneragan, 1983), or to growth setbacks associated with either stem, root or crown damage (Kellas et al., 1984; Landsberg, 1994; Murphy et al., 2010). Assessment of the former is the subject of our ongoing research. Evidence in support of the latter includes significant decreases in bark thickness of E. obliqua in fire treatments (increasing the potential for damage of cambial tissue; Gill and Ashton, 1968), and greatest growthdecreases in stems less than 50 cm dbhub, which would be more proportionally affected by scorching than larger, taller stems (Kellas et al., 1984). Growth of individual trees in this study was potentially influenced by climatic conditions over the study’s 27 years, and by changes in stand structure. Much of southern Australia was affected by a twelve-year drought prior to 2010 (Nicholls and Larsen, 2001), so any negative effects of fire on tree growth might have been exacerbated by low water availability for part of the measurement period (Keeling and Sala, 2012). Our frequency of diameter measures did not allow for an assessment of fire treatment by drought interactions, but we note that climate change projections indicate more frequent droughts for southern Australia (CSIRO, 2007) suggesting this study’s growth responses will have relevance to prescribed fire/climate interactions in the future. High mean mortality of small trees (c. 30–60%) might also have influenced diameter increments of some large trees by reducing competition and thereby ‘releasing’ growth (Hurteau and North, 2009). However, from both our plot-based and individual tree measures, we found no evidence of greater mortality of small trees in prescribed fire than control treatments. Combined with the above-mentioned low mortality of large trees, this suggested that effects of prescribed fire on tree growth in this eucalypt forest were not clearly moderated by changes in stand structure as has been observed in other forest types (Hurteau and North, 2009; Blanck et al., 2013). 4.3. Prescribed fire and tree-based carbon stocks Our study indicated mean decreases in total standing treebased carbon stocks with repeated prescribed fire of 25 Mg ha1, or about 12% of control tree-based stocks. However, the wide 95% confidence interval for this mean difference (2–48 Mg ha1), indicated a high degree of uncertainty around the size of the prescribed fire effect, which is not surprising given the level of treatment replication (five), and the variable structure of these native forests. Comparable empirical data of prescribed fire effects for Australian temperate forests are lacking, although a 12% decrease is comparable with decreases in tree-based carbon stocks of 11– 17% after single prescribed fires in the mixed-conifer forests of western USA (Meigs et al., 2009; North et al., 2009). Our finding of decreases in standing tree-based carbon stocks contradicts recent long-term simulation models that found no effect of prescribed fire regimes on carbon stocks in temperate eucalypt forests of south-eastern Australia (Norris et al., 2010; King et al., 2011), which was explained by quick recovery of stocks to pre-fire levels, and by ‘a limited impact’ of low intensity fires on tree biomass and growth (Norris et al., 2010). Decreases in standing tree-based carbon stocks associated with prescribed fire in this study appeared to be due to both direct and indirect effects. Direct combustion effects were low in bark (0.2– 0.4 Mg ha1), and were not detected (as a difference between control and fire treatments) in stumps; although decreased stump carbon with more frequent combustion was indicated by significantly lower stump carbon stocks in High than Low frequency fire treat-

253

ments (c. 1 t ha1). Direct combustion effects were also indicated by significantly less carbon in large dead stems in prescribed fire than control treatments despite higher mortality of large individual trees in the former. It’s also possible that repeated fire led to the collapse of some large stems through the formation of basal scars (Collins et al., 2012). These would be more likely after the higher severity fires in the Low than High frequency treatment (Gill, 1974), perhaps contributing to the lowest large dead stem carbon stocks in the Low frequency treatment. It is conceivable that decreases in standing tree-based carbon stocks in prescribed fire treatments could, through tree collapse, have been retained on site as (fallen) coarse woody debris. However, while not considered explicitly here, the following suggests this was not the case: (1) recent measures of carbon stocks before and after single prescribed fires in E. obliqua forest indicate a net loss of carbon stocks in fallen timber (i.e. combustion losses greater than tree-fall gains; Volkova and Weston, 2013); (2) this net loss finding is consistent with our preliminary (unpublished) assessments of coarse woody debris in the study plots that indicate greater losses through repeated combustion than gains through tree collapse; and (3) low overall rates of large eucalypt stem collapse after low intensity prescribed fire are indicated by low rates of collapse (6 0.2% year1) after post-logging fires of greater intensity in comparable eucalypt forests (Gibbons et al., 2000). In addition to direct combustion, indirect effects of prescribed fire treatments on tree-based carbon stocks were detected in both marginally increased post-fire mortality and decreased mean annual diameter increment of large E. obliqua. Significant decreases in the mean diameter increment of the most prevalent large E. obliqua size class (30 to 50 cm dbhub) represented a decrease in tree carbon mass of about 0.1 Mg ha1 year1. The net ecosystem exchange (NEE) of carbon in our study’s forest is the subject of ongoing research, but, based on values of 7 Mg C ha1 year1 in wetter eucalypt forests (Keith et al., 2012), might conservatively be in the range 3–5 Mg C ha1 yr1 (see also Beringer et al., 2007). Thus, a decrease of 0.1 Mg ha1 year1 might seem negligible, although we note that a decrease of 0.2 Mg C ha1 year1 in northern Australian savannas after moderate fires was considered ‘relatively large’ for an NEE of 3–5 Mg C ha1 yr1 (Murphy et al., 2010). In addition, while the deep-barked E. obliqua was the most common eucalypt, it contained just 41% of the large live tree carbon in control plots, with the remainder comprised of two thinner-barked species (E. radiata, 33%; E. rubida, 26%; data not shown). Thus, while limited (non-replicated) data from one of our study areas indicated no negative effects of prescribed fire on individual E. rubida growth, we cannot discount the possibility that repeated prescribed fire impacted on the growth of E. radiata and E. rubida, thereby contributing to an overall finding of less tree-based carbon stocks in prescribed fire than control treatments. While we found no significant differences among our prescribed fire treatments in total standing tree-based carbon stocks, analyses of component stocks indicated that both fire season and frequency could influence the capacity of trees to fix carbon into the future. For example, recruitment and survival of smaller stems was greater after autumn than spring fires suggesting that autumn fires would better maintain future growth stocks. Similarly, small tree recruitment and overall large tree survival were greater after Low than High frequency fires, suggesting that 10- rather than 3yearly fires would better maintain the forest’s future capacity to fix carbon.

5. Conclusions In a first for temperate eucalypt forests, we detected significant effects of prescribed fire treatments on tree mortality and growth,

254

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255

and on total standing tree-based carbon stocks. That such effects have not previously been reported is evidence of the importance of long-term studies, examining the cumulative effects of multiple prescribed fires, and of using multiple data sources that address both current stocks and long-term fluxes (mortality and growth). We found evidence of both direct and indirect effects of prescribed fire on standing tree-based carbon stocks. That is, direct combustion effects were indicated by minor bark losses, and by significantly less carbon in large dead stems in prescribed fire than control treatments. Indirect effects included marginally greater mortality and decreased mean diameter increment of individual large trees in prescribed fire than control treatments over the 27 years since treatment establishment. These direct and indirect effects were consistent with a detected mean difference in total standing tree-based carbon stocks of 25 Mg ha1 between control and prescribed fire treatments. However, the wide confidence interval for this difference (2–48 Mg ha1) indicated that a high degree of uncertainty remains about the magnitude of decreases in standing tree-based carbon stocks under prescribed fire regimes in these native eucalypt forests. Our analyses of prescribed fire types did not detect significant effects of fire season and frequency on total standing tree-based carbon stocks. However, densities of small live stems and associated carbon stocks were greater in autumn than spring, and in Low than High frequency fire treatments. Combined with lower plot-level proportions of dead stems in Low than High frequency treatments, our study indicates that autumn fires every c. 10 years would offer more scope than more frequent spring fires to maintain the capacity of these forests to fix carbon into the future. Acknowledgements This work was funded by the Victorian Department of Environment and Primary Industries (DEPI). We thank many personnel from DEPI’s (then) Environmental Policy and Climate Change, Land and Fire, and Forests and Parks Divisions for supporting this FESA measurement. We also thank many current and past staff from the regional DEPI and from the University of Melbourne for maintaining, measuring, and documenting the prescribed fire treatments for over two decades, including J. Kellas, D. Oswin, A. Ashton, J. Najera, and N. Klaus. Two anonymous reviewers are also thanked for providing constructive comments that improved the paper. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.foreco.2013.06.036. References Abbott, I., Loneragan, O., 1983. Influence of fire on growth rate, mortality, and butt damage in Mediterranean forest of Western Australia. For. Ecol. Manage. 6, 139–153. Abbott, I., Loneragan, O., 1984. Growth rate and long-term population dynamics of jarrah (Eucalyptus marginata Donn ex Sm.) regeneration in Western Australian forest. Aust. J. Bot. 32, 353–362. Adams, M.A., 2013. Mega-fires, tipping points and ecosystem services: managing forests and woodlands in an uncertain future. For. Ecol. Manage. 294, 250–261. Adams, M.A., Attiwill, P.M., 2011. Burning Issues: Sustainability and Management of Australia’s Southern Forests. CSIRO Publishing, Collingwood, Victoria, Australia. ASRIS, 2011. ASRIS - Australian Soil Resource Information System. http:// www.asris.csiro.au. Accessed: April 7, 2013. Attiwill, P.M., 1979. Nutrient cycling in a Eucalyptus obliqua (L’Herit.) forest. III. Growth, biomass, and net primary production. Aust. J. Bot. 27, 439–458. Attiwill, P.M., Adams, M.A., 2008. Harnessing forest ecological sciences in the service of stewardship and sustainability: a perspective from ‘down-under’. For. Ecol. Manage. 256, 1636–1645.

Benyon, R.G., Lane, P.N.J., 2013. Ground and satellite-based assessments of wet eucalypt forest survival and regeneration for predicting long-term hydrological responses to a large wildfire. For. Ecol. Manage. 294, 197–207. Beringer, J., Hutley, L.B., Tapper, N.J., Cernusak, L.A., 2007. Savanna fires and their impact on net ecosystem productivity in North Australia. Global Change Biol. 13, 990–1004. Bi, H., Turner, J., Lambert, M.J., 2004. Additive biomass equations for native eucalypt forest trees of temperate Australia. Trees – Struct. Funct. 18, 467–479. Blanck, Y.-L., Rolstad, J., Storaunet, K.O., 2013. Low- to moderate-severity historical fires promoted high tree growth in a boreal Scots pine forest of Norway. Scand. J. For. Res. 28, 126–135. Bradstock, R.A., 2010. A biogeographic model of fire regimes in Australia: current and future implications. Global Ecol. Biogeogr. 19, 145–158. Bradstock, R.A., Williams, R.J., 2009. Can Australian fire regimes be managed for carbon benefits? New Phytol. 183, 931–934. Bradstock, R.A., Boer, M.M., Cary, G.J., Price, O.F., Williams, R.J., Barrett, D., Cook, G., Gill, A.M., Hutley, L.B.W., Keith, H., Maier, S.W., Meyer, M., Roxburgh, S.H., Russell-Smith, J., 2012. Modelling the potential for prescribed burning to mitigate carbon emissions from wildfires in fire-prone forests of Australia. Int. J. Wildland Fire 21, 629–639. Burrows, N., Ward, B., Robinson, A., 2010. Fire regimes and tree growth in low rainfall Jarrah forest of south-west Australia. Environ. Manage. 45, 1332– 1343. Busse, M.D., Simon, S.A., Riegel, G.M., 2000. Tree-growth and understory responses to low-severity prescribed burning in thinned Pinus ponderosa forests of central Oregon. For. Sci. 46, 258–268. Campbell, J.L., Harmon, M.E., Mitchell, S.R., 2011. Can fuel-reduction treatments really increase forest carbon storage in the western US by reducing future fire emissions? Front. Ecol. Environ. 10, 83–90. Carter, M.C., Darwin Foster, C., 2004. Prescribed burning and productivity in southern pine forests: a review. For. Ecol. Manage. 191, 93–109. Chatto, K., Bell, T.L., Kellas, J., 2003. Effects of repeated low-intensity fire on tree growth and bark in a mixed eucalypt foothill forest in south-eastern Australia. Fire Research Report No. 66. Fire Management, Department of Sustainability and Environment, East Melbourne, Victoria, Australia. Clarke, H.G., Smith, P.L., Pitman, A.J., 2011. Regional signatures of future fire weather over eastern Australia from global climate models. Int. J. Wildland Fire 20, 550–562. Collins, L., Bradstock, R.A., Tasker, E.M., Whelan, R.J., 2012. Can gullies preserve complex forest structure in frequently burnt landscapes? Biol. Conserv. 153, 177–186. Coomes, D.A., Allen, R.B., Scott, N.A., Goulding, C., Beets, P., 2002. Designing systems to monitor carbon stocks in forests and shrublands. For. Ecol. Manage. 164, 89– 108. CSIRO, 2007. Climate change in Australia: Technical Report 2007. CSIRO Marine and Atmospheric Research, Aspendale, Victoria, Australia. DSE, 2003. Ecological impacts of fuel reduction burning in a mixed eucalypt foothill forest – summary report (1984–1999). Fire Research Report No. 57. Department of Sustainability and Environment, East Melbourne, Victoria, Australia. DSE, 2012. Code of Practice for Bushfire Management on Public Land. Department of Sustainability and Environment, East Melbourne, Victoria, Australia. Fensham, R.J., Fairfax, R.J., Buckley, Y.M., 2008. An experimental study of fire and moisture stress on the survivorship of savanna eucalypt seedlings. Aust. J. Bot. 56, 693–697. Fernandes, P.M., Botelho, H.S., 2003. A review of prescribed burning effectiveness in fire hazard reduction. Int. J. Wildland Fire 12, 117–128. Flannigan, M., Cantin, A.S., de Groot, W.J., Wotton, M., Newbery, A., Gowman, L.M., 2013. Global wildland fire season severity in the 21st century. For. Ecol. Manage. 294, 54–61. Floyd, A.G., 1966. The effects of control burning on forests. In: The effects of fire on forest conditions. Technical Paper No. 13. Forestry Commission of New South Wales, Taree, New South Wales, Australia, pp. 3–20. Ghimire, B., Williams, C.A., Collatz, G.J., Vanderhoof, M., 2012. Fire-induced carbon emissions and regrowth uptake in western U.S. forests: documenting variation across forest types, fire severity, and climate regions. J. Geophys. Res.: Biogeosci. 117, 1–29. Gibbons, P., Lindenmayer, D.B., Barry, S.C., Tanton, M.T., 2000. The effects of slash burning on the mortality and collapse of trees retained on logged sites in southeastern Australia. For. Ecol. Manage. 139, 51–61. Gill, A.M., 1974. Towards an understanding of fire scar formation: field observation and laboratory simulation. For. Sci. 20, 198–205. Gill, A.M., 1980. Restoration of bark thickness afer fire and mechanical injury in a smooth-barked eucalypt. Austral. For. Res. 10, 311–319. Gill, A.M., Ashton, D.H., 1968. The role of bark type in relative tolerance to fire of three Central Victorian eucalypts. Aust. J. Bot. 16, 491–498. Gill, A.M., Brack, C.L., Hall, T., 1982. Bark probe – an instrument for measuring bark thickness of eucalypts. Austral. For. 45, 206–208. Grierson, P., Adams, M., Attiwill, P., 1992. Estimates of carbon storage in the aboveground biomass of Victoria’s forests. Aust. J. Bot. 40, 631–640. Grove, S.J., Stamm, L., Barry, C., 2009. Log decomposition rates in Tasmanian Eucalyptus obliqua determined using an indirect chronosequence approach. For. Ecol. Manage. 258, 389–397. Guinto, D.F., House, A.P.N., Xu, Z.H., Saffigna, P.G., 1999. Impacts of repeated fuel reduction burning on tree growth, mortality and recruitment in mixed species eucalypt forests of southeast Queensland. Austral. For. Ecol. Manage. 115, 13– 27.

L.T. Bennett et al. / Forest Ecology and Management 306 (2013) 243–255 Hurteau, M., North, M., 2009. Fuel treatment effects on tree-based forest carbon storage and emissions under modeled wildfire scenarios. Front. Ecol. Environ. 7, 409–414. Kaye, J., 2009. Estimation of tree above-ground biomass across native forests of Victoria. Masters by Research Thesis. Melbourne School of Land and Environment, The University of Melbourne, Victoria, Australia Keeling, E.G., Sala, A., 2012. Changing growth response to wildfire in old-growth ponderosa pine trees in montane forests of north central Idaho. Global Change Biol. 18, 1117–1126. Keith, H., Barrett, D., Keenan, R., 2000. Review of allometric relationships for estimating woody biomass for New South Wales, the Australian Capital Territory, Victoria, Tasmania and South Australia. National Carbon Accounting Technical Report, vol. 5B. Australian Greenhouse Office, Canberra, ACT, Australia. Keith, H., van Gorsel, E., Jacobsen, K.L., Cleugh, H.A., 2012. Dynamics of carbon exchange in a Eucalyptus forest in response to interacting disturbance factors. Agric. For. Meteorol. 153, 67–81. Kellas, J.D., Incoll, W.D., Squire, R.O., 1984. Reduction in basal area increment of Eucalyptus obliqua following crown scorch. Austral. For. 47, 179–183. King, K.J., de Ligt, R.M., Cary, G.J., 2011. Fire and carbon dynamics under climate change in south-eastern Australia: insights from FullCAM and FIRESCAPE modelling. Int. J. Wildland Fire 20, 563–577. King, K.J., Cary, G.J., Bradstock, R.A., Marsden-Smedley, J.B., 2013. Contrasting fire responses to climate and management: insights from two Australian ecosystems. Global Change Biol. 19, 1223–1235. Landsberg, J.D., 1994. A review of prescribed fire and tree growth response in the genus Pinus. In: Proceedings of the 12th Conference on Fire and Forest Meteorology. Society of American Foresters, Bethesda, MD, pp. 326–349. Lawes, M.J., Midgley, J.J., Clarke, P.J., 2013. Costs and benefits of relative bark thickness in relation to fire damage: a savanna/forest contrast. J. Ecol. 101, 517– 524. Luke, R.H., McArthur, A.G., 1978. Bushfires in Australia. Australian Government Publishing Service, Canberra, Australia. McArthur, A.G., 1967. Fire behaviour in eucalypt forests. Leaflet No. 107. Forest Research Institute, Forestry and Timber Bureau of Australia, Canberra, Australia. Meigs, G., Donato, D., Campbell, J., Martin, J., Law, B.E., 2009. Forest fire impacts on carbon uptake, storage, and emission: the role of burn severity in the Eastern Cascades, Oregon. Ecosystems 12, 1246–1267. Moroni, M.T., 2012. Aspects of forest carbon management in Australia: a discussion paper. For. Ecol. Manage. 275, 111–116. Mount, A.B., 1972. The Derivation and Testing of a Soil Dryness Index using Run-off Data. Tasmanian Forestry Commission Bulletin No. 4. Tasmanian Forestry Commission, Hobart, Tasmania, Australia. Murphy, B.P., Russell-Smith, J., Prior, L.D., 2010. Frequent fires reduce tree growth in northern Australian savannas: implications for tree demography and carbon sequestration. Global Change Biol. 16, 331–343. Narayan, C., Fernandes, P.M., van Brusselen, J., Schuck, A., 2007. Potential for CO2 emissions mitigation in Europe through prescribed burning in the context of the Kyoto Protocol. For. Ecol. Manage. 251, 164–173. Nicholls, N., Larsen, S., 2001. Impact of drought on temperature extremes in Melbourne, Australia. Austral. Meteorol. Ocean. J. 61, 113–116. Norris, J., Arnold, S., Fairman, T., 2010. An indicative estimate of carbon stocks on Victoria’s publicly managed land using the FullCAM carbon accounting model. Austral. For. 73, 209–219. North, M.P., Hurteau, M.D., 2011. High-severity wildfire effects on carbon stocks and emissions in fuels treated and untreated forest. For. Ecol. Manage. 261, 1115–1120.

255

North, M., Hurteau, M., Innes, J., 2009. Fire suppression and fuels treatment effects on mixed-conifer carbon stocks and emissions. Ecol. Appl. 19, 1385–1396. Opie, J.E., 1968. Predictability of individual tree growth using various definitions of competing basal area. For. Sci. 14, 314–323. Parliament of Victoria, 2010. 2009 Victorian Bushfires Royal Commission: Final Report. The Commission, Melbourne, Victoria, Australia. Payne, R., 2011. A Guide to Anova and Design in GenStatÒ. VSN International, Hemel Hempstead, Hertfordshire, UK. Penman, T.D., Christie, F.J., Andersen, A.N., Bradstock, R.A., Cary, G.J., Henderson, M.K., Price, O., Tran, C., Wardle, G.M., Williams, R.J., York, A., 2011. Prescribed burning: how can it work to conserve the things we value? Int. J. Wildland Fire 20, 721–733. Peterson, D.W., Reich, P.B., 2001. Prescribed fire in oak savanna: fire frequency effects on stand structure and dynamics. Ecol. Appl. 11, 914–927. Quinn, G.P., Keough, M.J., 2002. Experimental Design and Data Analysis for Biologists. Cambridge University Press, Cambridge, UK. Rowan, J.N., Russell, L.D., Ransom, S.W., Rees, D.B., 2000. Land Systems of Victoria (third ed.). Technical Report No. 56. Centre for Land Protection Research, Department of Natural Resources and Environment, Melbourne, Victoria, Australia. San-Miguel-Ayanz, J., Moreno, J.M., Camia, A., 2013. Analysis of large fires in European Mediterranean landscapes: lessons learned and perspectives. For. Ecol. Manage. 294, 11–22. Specht, R.L., 1981. Foliage projective cover and standing biomass. In: Gillson, A.N., Anderson, D.J. (Eds.), Vegetation Classification in Australia. CSIRO, Canberra, Australia, pp. 10–21. Speight, J.G., 2009. Landform. In, Australian Soil and Land Survey Field Handbook (third ed.). The National Committee on Soil and Terrain, CSIRO Publishing, Collingwood, Victoria, Australia, pp. 15–72. Strasser, M.J., Pausas, J.G., Noble, I.R., 1996. Modelling the response of eucalypts to fire, Brindabella ranges, ACT. Aust. J. Ecol. 21, 341–344. Tolhurst, K.G., 2003. Effects of repeated low-intensity fire on the understorey of a mixed eucalypt foothill forest in south-eastern Australia. Fire Research Report No. 58. Fire Management, Department of Sustainability and Environment, East Melbourne, Victoria, Australia. Tolhurst, K.G., Flinn, D., 1992. Ecological impacts of fuel reduction burning in dry sclerophyll forest: first progress report. Forest Research Report No. 349. Forest Research Section, Department of Conservation and Environment, Kew, Victoria, Australia. van Mantgem, P.J., Stephenson, N.L., Knapp, E., Battles, J., Keeley, J.E., 2011. Longterm effects of prescribed fire on mixed conifer forest structure in the Sierra Nevada, California. For. Ecol. Manage. 261, 989–994. van Mantgem, P.J., Nesmith, J.C.B., Keifer, M., Brooks, M., 2013. Tree mortality patterns following prescribed fire for Pinus and Abies across the southwestern United States. For. Ecol. Manage. 289, 463–469. Vilén, T., Fernandes, P., 2011. Forest fires in Mediterranean countries: CO2 emissions and mitigation possibilities through prescribed burning. Environ. Manage. 48, 558–567. Volkova, L., Weston, C., 2013. Redistribution and emission of forest carbon by planned burning in Eucalyptus obliqua (L. Hérit.) forest of south-eastern Australia. For. Ecol. Manage., . Warton, D.I., Hui, F.K.C., 2011. The arcsine is asinine: the analysis of proportions in ecology. Ecology 91, 3–10. Wilkinson, G., Jennings, S., 1993. Survival and recovery of Eucalytpus obliqua regeneration following wildfire. TasForests 5, 1–11.