Volume 14, Issue 2 e12766
LETTER
Open Access

Does forest thinning reduce fire severity in Australian eucalypt forests?

Chris Taylor

Chris Taylor

Fenner School of Environment and Society, The Australian National University, Canberra, Australian Capital Territory, Australia

Search for more papers by this author
Wade Blanchard

Wade Blanchard

Fenner School of Environment and Society, The Australian National University, Canberra, Australian Capital Territory, Australia

Search for more papers by this author
David B. Lindenmayer

Corresponding Author

David B. Lindenmayer

Fenner School of Environment and Society, The Australian National University, Canberra, Australian Capital Territory, Australia

Correspondence

David B. Lindenmayer, Fenner School of Environment and Society, The Australian National University, 141 Linnaeus Way, Canberra, ACT 2601, Australia.

Email: [email protected]

Search for more papers by this author
First published: 15 September 2020
Citations: 13

Abstract

Forest thinning has been proposed to reduce fire severity. However, evidence to support this strategy in Australia is scant. We completed a detailed empirical analysis of stand history data from forests burned in wildfires in 2009 in south-eastern Australia, to address the question: Does forest thinning reduce fire severity? The answer varied depending on fire type (Crown Burn vs. Crown Burn/Crown Scorch), forest type, and stand age. For the statistical relationship for Crown Burn, there were no thinning effects in ash-type forests. For mixed species forests, thinning reduced the probability of Crown Burn in young stands but increased it in older stands. Data for the fire severity category of Crown Burn/Crown Scorch revealed that thinning generally elevated fire severity, irrespective of stand age, forest type, or fire zone. Except for 20- to 40-year-old mixed species forest subject to Crown Burn, proposals for thinning to reduce fire severity have limited support.

1 INTRODUCTION

Fire is a key factor influencing ecosystem structure, condition, composition, and processes (Bowman et al., 2009) with large-scale, high-severity wildfires presenting major challenges to policy makers and resource managers globally. Climate and weather are the major drivers of fire regimes (Jones et al., 2020). However, other factors can influence the occurrence and severity of wildfires, including the amount and/or age of fire fuel (Bradstock, Gill, & Williams, 2012), the prevalence of human ignitions (Collins, Price, & Penman, 2015), and past land management, such as forestry operations (Taylor, McCarthy, & Lindenmayer, 2014).

There have been extensive wildfires worldwide in the past few years, including in the Amazon, western North America, the Arctic, and Australia. There has been much public discussion in Australia about ways to reduce the impacts of fire, including reducing fire severity across the broader forest estate. Based on computer simulations, Volkova, Bia, Hilton, and Weston (2017) suggested that thinning can reduce fire severity in Australian eucalypt forests. Therefore, Australian forest industry advocates have pushed for a substantial expansion of commercial thinning operations to reduce the risk of high severity wildfire (AFPA, 2020). Thinning of inland conifer forests of south-western USA may restore “natural” fire regimes less damaging to particular vegetation types (Friederici, 2003). However, in an Australian context, there is evidence that thinning may increase wildfire risks in eucalypt forests (Buckley & Corkish, 1991; LaSala, 2001). Given these conflicting perspectives, it is critical to conduct empirical, field-based analyses to quantify relationships between interventions, such as thinning and other logging operations, and the severity of subsequent wildfires.

We posed the question: Does forest thinning reduce fire severity? To answer this question, we examined empirical field data ash-type forest (Mountain Ash [Eucalyptus regnans] and Alpine Ash [Eucalyptus delegatensis]) and mixed eucalypt species forests burned in the 2009 wildfires across the Central Highlands of Victoria, south-eastern Australia. We compared fire severity (sensu Keeley, 2009) in forests that had been thinned versus those that remained unthinned. We examined two classes of fire severity, one in which forests were subject to a Crown Burn (where the leaves were consumed in a fire) and those subject to Crown Scorch (where the leaves were scorched but not consumed by a fire burning below the crown). More specifically, we tested a series of hypotheses related to thinning:
  • H0: A null response in which there were no relationships between thinning and fire severity.
  • H1: A negative relationship in which thinning reduced fire severity as suggested from simulation studies (Volkova et al., 2017).
  • H2: A positive relationship in which thinning resulted in an increased severity as indicated from other work on thinning in eucalypt forests (LaSala, 2001).
We also sought to determine if the effects of thinning on fire severity varied in response to other factors, such as stand age, forest type, and the weather at the time of fire. Hence, we tested for three-way interactions in which the effects of thinning would be different across the levels of the other factors. We hypothesized that the effects of thinning on fire severity would be:
  • More pronounced in younger forests than older stands. This was because previous studies have shown that fire severity is inherently lower in older forests (Taylor et al., 2014). Hence, thinning will have less impact in forests where fire severity is already reduced.
  • More pronounced in ash-type forests than mixed species forests. We made this prediction because ash-type forests are higher biomass ecosystems than drier, mixed species forests. Thinning may, therefore, have relatively greater suppressive effects on fire severity in ash-type forests. And/or;
  • More muted under extreme fire conditions. We hypothesized this response because of the well-documented effects of extreme weather on fire behavior (e.g., see Sullivan, McCaw, Cruz, Matthews, & Ellis, 2012). Thus, the potential effectiveness of interventions, such as thinning, may manifest only under low to moderate fire weather and not under extreme fire weather.

2 METHODS

2.1 Study area

Our investigation focused on 227,560 ha of forest burned across the Central Highlands of Victoria (south eastern Australia) (Figure 1) in the 2009 “Black Saturday” wildfires. The study area is dominated largely by wet sclerophyll forests dominated by Mountain Ash or Alpine Ash (collectively referred hereafter as ash-type forests), along with mixed species forests, such as those supporting Messmate (Eucalyptus obliqua), Mountain Grey Gum (Eucalyptus cypellocarpa), and Manna Gum (Eucalyptus viminalis) trees. The main form of disturbance in these forests is wildfire, which typically occurs during dry summer conditions, particularly periods of prolonged drought.

Details are in the caption following the image
Location of study area

Our study area was disturbed by two wildfires, the Kilmore East and the Murrindindi fires in February 2009, ignited at 11:45 and 14:55 hr, respectively, on February 7, 2009 (Gellie, Mattingley, Gibos, Wells, & Salkin, 2013). Following a wind change occurring between 18.00 and 19.00 hr on that day, these two wildfires joined into a single large wildfire complex (Gellie et al., 2013).

The burned forests in our study area have also been subject to widespread clearcutting, which involves the removal of all merchantable logs in one integrated logging operation. Remaining slash and forest debris is then burned in a high-intensity planned burn (Flint & Fagg, 2007). Stands of ash-type and mixed species forest have been thinned in aggregated cutblocks, with some of those stands regrowing following past clearcutting operations (DJPR, 2019) (see Supporting Information––Summary S1).

2.2 Fire weather

We divided our study area into two fire zones (Figure 2) corresponding to severe fire weather (Zone 1) and moderate to low fire weather (Zone 2) (see Summary S1 for further details of fire weather zones). We used two broad zones to sample sufficient numbers of sites under each zone that had been thinned across our study area (Table S1).

Details are in the caption following the image
Fire zones (corresponding to the progression of the fire across our study area), forest types, fire severity (see text), and forest age classes at the time of the 2009 wildfires

2.3 Fire severity and other spatial data

We sampled data from a series of Victorian Government data layers to assemble spatial data and satellite information on forest-type cover, logging history, and the severity of the fires in 2009 (Figure 2). Fire severity was categorized across five classes ranging from a canopy-consuming fire (the most severe) to unburned (the lowest) (see Summary S1).

2.4 Data sampling

We sampled fire severity data using a square grid of site points at 500 m intervals for unthinned sites and 100 m intervals for thinned sites across our study area (Figure 3) (see Summary S1). We identified 32 cutblocks commercially thinned prior to the 2009 wildfires (Figure 4) on slopes 0-10 degrees, covering an area of 564 ha. In total, we obtained data for 424 thinned sites and 1,687 unthinned sites (Table S2). Details of the data sampling methods are given in Summary S2.

Details are in the caption following the image
Sample points for thinned (yellow dots) and unthinned sites in the study area
Details are in the caption following the image
Ash forest unthinned, thinned, and thinned followed by wildfire. Photograph (a) is of unthinned Alpine Ash Forest near our study area around Mount Baw Baw; Photograph (b) is of Alpine Ash forest on the Blue Range in our study area that was subject to thinning prior to the February 2009 wildfires; Photograph (c) is of the same Alpine Ash forest in Photograph B following the February 2009 wildfires

2.5 Statistical analysis

We employed Bayesian logistic regression to assess which factors were associated with our two fire severity response variables: (1) the probability of Crown Burn and (2) the probability of the combined category of Crown Burn/Crown Scorch. The five potential explanatory variables we examined were: thinning (unthinned vs. thinned), forest type (ash-type forest vs. mixed species forest), forest age (20–40 vs. 70 years old), fire zone corresponding to the weather conditions at the time a site was burned in 2009 (extreme conditions [Zone 1] or moderate conditions [Zone 2]), and whether a site was facing the approaching fire (no or yes) (Table S4). Further details of the statistical analysis for this study are given in Summary S2.

3 RESULTS

The area of thinned forest sustaining a Crown Burn was 252 ha or 31% of the total area. The area sustaining a Crown Burn/Crown Scorch was 531 ha or 66% of the total area thinned within the wildfire extent. The percentage of sites that experienced Crown Burn and Crown Burn/Crown Scorch broken down by our five factors is given in Table S5. We present the results for models for Crown Burn and Crown Burn/Crown Scorch, separately (Tables S6 and S7). The adequacy of our best fitting models as measured by AUC ROC was 0.83 and 0.95 for Crown Burn and Crown Burn/Crown Scorch, respectively (Table S8). The model fit for Crown Burn (Nagelkerke R= 0.367) was lower than it was for Crown Burn/Crown Scorch (Nagelkerke R= 0.757) (Table S8).

3.1 Thinning effects

The best fitting model for the fire severity response variable of Crown Burn contained a three-way interaction between thinning, stand age, and forest type (Figure 5a, Tables S6 and S7, and Figure S2). For 70-year-old mixed species forests, the probability of Crown Burn was higher in thinned forest than unthinned forest (log odds ratio [LOR] = 1.09, 95% credible interval [CI] [0.73, 1.48]). This effect was reversed in 20- to 40-year-old mixed species forest (LOR = –1.42, 95% CI [–2.25, –0.65]). There was no evidence of thinning effects in ash-type forests, irrespective of stand age (LOR = 0.04 95% CI [–1.48, 1.37] and LOR = 0.09, 95% CI [–0.71, 0.91] for 70-year-old and 20- to 40-year-old ash forest, respectively) (Figures 5a and S2). We found marginal evidence of an interaction effect between fire conditions and thinning; the probability of a Crown Burn was marginally lower in thinned areas under moderate fire conditions (Zone 2) (LOR = –2.88, 95% CI [–7.91, 0.12]) and there was no difference in probability of a Crown Burn under extreme fire conditions (Zone 1) (LOR = 0.09, 95% CI [–0.71, 0.91]) (Figures 5b and S2).

Details are in the caption following the image
Factors that interact with thinning to influence the probability of either a Crown Burn or a Crown Burn/Crown Scorch. Thinned is shown in grey and unthinned is shown in brown. Panel a: Estimated probability of a Crown Burn showing the three-way interaction of forest type, forest age, and thinning for a zone 1 fire not facing the approaching fire. Panel b: Estimated probability of a Crown Burn showing the interaction between fire zone and thinning for 20- to 40-year-old ash forest not facing the approaching fire. Panel c: Estimated probability of Crown Burn/Crown Scorch showing the three-way interaction between forest age, fire zone, and thinning for an ash forest not facing the approaching fire. The error bars denote 95% credible intervals. Significant and marginally significant comparisons between thinned and nonthinned stands are indicated by different letters for the appropriate comparison. Where the letters for a given comparison are the same, there is no evidence of a difference. See Figures S2 and S3 for selected pairwise comparisons of the factor levels and their interactions on log odds ratio scale

The best fitting model for the factors influencing the composite fire severity response variable of Crown Burn/Crown Scorch included a three-way interaction between thinning, stand age, and zone (Figure 5c, Tables S6 and S8, and Figure S3). The probability of Crown Burn/Crown Scorch was marginally elevated by thinning in younger (20- to 40-year-old) forests under extreme fire weather conditions (Zone 1) (LOR = 0.98, 95% CI [–0.05, 2.13]) and significantly elevated under moderate to low fire weather conditions (Zone 2) (LOR = 3.08, 95% CI [1.24, 5.72]). Thinning significantly increased the probability of Crown Burn/Crown Scorch in 70-year-old forest burned under extreme fire weather conditions (Zone 1) (LOR = 1.48, 95% CI [0.14, 3.02]), but there was no change for 70-year-old forest burned under moderate to low fire weather conditions (Zone 2) (LOR = 0.14, 95% CI [–1.06, 1.12]) (Figures 5c and S3).

3.2 Effects on fire severity of stand age, forest type, zone, and fire direction

The probability of Crown Burn was lower in 70-year-old unthinned stands than in younger (20- to 40-year-old) unthinned stands across both ash-type forest and mixed species forest (LOR = –1.55, 95% CI [–2.18, –0.96], LOR = –0.77, 95% CI [–1.33, .–021] for ash and mixed species unthinned stands, respectively) (Figures 5a and S2). A similar stand age effect was found for thinned ash forest (LOR = –1.66, 95% [–3.18, –0.22]); however, the effect was reversed for thinned mixed species forest (LOR = 1.75, 95% CI [1.10, 2.44]) (Figures 5a and S2).

The probability of a Crown Burn was lower in forests burned under moderate versus extreme fire weather conditions (i.e., Zone 1 vs. Zone 2) in unthinned and thinned forests (LOR = –4.12, LOR [–5.11, –3.3] and LOR = 7.11, 95% [–12.15, –4.24] for unthinned and thinned forests, respectively) (Figures 5b and S2). There was a marginal increase in the probability of Crown Burn on forests facing the approaching fire compared with forests not facing the approaching fire (LOR = 0.31, 95% CI [0.02, 0.59]) (Table S6, Figures S1 and S3).

The probability of Crown Burn/Crown Scorch was lower in forests under low and moderate conditions (Zone 2) relative to areas burned under extreme conditions (Zone 1) irrespective of forest age or thinning (20- to 40-year-old, unthinned LOR = –6.02, 95% CI [–7.85, –.4.43], 20- to 40-year-old, thinned LOR = –5.53, 95% CI [–7.09, –4.19], 70-year-old, unthinned LOR = –5.71, 95% CI [–6.42, –5.07], and 70-year-old, thinned LOR = –8.65, 95% CI [–11.65, –6.49]) (Figures 5c and S3, Table S7). Other effects for the composite fire severity response variable of Crown Burn/Crown Scorch included: (1) A forest age × forest-type interaction in which the probability of Crown Burn/Crown Scorch was higher in 70-year-old forest relative to 20- to 40-year-old stands but only for mixed species forests (LOR = 1.37, 95% CI [0.49, 2.24]), with no such effect in ash-type forests (LOR = –0.04, 95% CI [–1.34, 1.11]) (Figures S1 and S3). (2) A stand age × fire direction interaction in which 70-year-old forests on sites facing the approaching fire had a lower probability of Crown Burn/Crown Scorch than younger (20- to 40-year-old) stands (LOR = –0.57, 95% CI [–1.00, –0.15]) (Figures S1 and S3). This was in contrast to sites not directly facing the approaching fire where there was a marginal increase (LOR = 0.78, 95% CI [–0.08, 1.64]) (Figures S1 and S3). And, (3) A forest type × fire zone interaction in which the difference in probability of a Crown Burn/Crown Scorch between moderate fire conditions (Zone 2) and extreme fire conditions (Zone 1) was greater in ash forest (LOR –6.02, 95% CI [–7.84, –4.42]) than mixed forest (LOR = –5.08, 95% CI [–7.01, –3.29]) (Figures S1 and S3).

4 DISCUSSION

Our key question was Does forest thinning reduce fire severity? The answer was complex as the effects of thinning varied with forest type, stand age, and fire conditions (i.e., zone). In our model of the factors influencing the probability of Crown Burn, there were no effects of thinning in ash-type forests. For mixed species forests, thinning effects were age-sensitive; it lowered the probability of Crown Burn in young stands but increased it in older stands (Figures 5a and S2). Our analysis of the composite Crown Burn/Crown Scorch fire severity response variable revealed that thinning either elevated fire severity or had no effect, irrespective of stand age, forest type, or fire zone (Figures 5c and S3).

Our findings for ash-type forests contrast with simulation modeling, which suggested that thinning reduced fire effects in Alpine Ash forest (Volkova et al., 2017). The reasons for these differences may be associated with differences between an empirical study (this investigation) compared with assumptions underpinning simulation modeling. Volkova et al. (2017) found that thinning operations led to an increase in elevated (ladder) fuels and in the model they used (see Cheney, Gould, McCaw, & Anderson, 2012); elevated fuels were the primary driver of flame height, which would likely have predicted greater flame heights. However, Volkova et al. (2017) did not consider elevated fuels in their conclusions and they may have obtained different relationships with thinning and fire severity if they had applied another model better suited to modeling these forests (see Zylstra et al., 2016).

We found that thinning can exacerbate the probability of Crown Burn (Figures 5a and S2) and Crown Burn/Crown Scorch (Figures 5c and S3). It is possible that although thinning removes trees from stands, such operations may also dry the forest floor due to both the loss of mesic understorey elements like tree ferns (which are sensitive to logging operations, Bowd, Lindenmayer, Banks, & Blair, 2018) and changes in microclimate, such as increased penetration of light to the forest floor (LaSala, 2001). Thinning may also increase air movement through a forest, thereby potentially facilitating the spread of fire through the forest. The debris left in the forest following thinning operations may also add fuels that promote increased fire severity (Buckley & Corkish, 1991).

4.1 Effects on fire severity of stand age, forest type, zone, and fire direction

The probability of a Crown Burn was markedly higher in forests burned under extreme versus moderate conditions (i.e., Zone 1 vs. Zone 2). This result is congruent with the widely documented effects of fire weather on fire behavior (e.g., see Sullivan et al., 2012). Thus, for example, when temperatures and wind speeds are high and relative humidity is low, thinning or indeed other prefire interventions (such as prescribed burning) may have only limited effects on reducing fire severity.

The probability of Crown Burn was greater in younger (20- to 40-year-old) stands of ash-type forest than in 70-year-old stands (Figure 5). This was consistent with the findings of earlier studies in these forests (Taylor et al., 2014) and other work demonstrating higher flammability of young forests relative to older stands elsewhere in Australia (Zylstra, 2018) and overseas (Tiribelli, Morales, Gowda, Mermoz, & Kitzberger, 2018; Zald & Dunn, 2017).

4.2 Caveats

We completed extensive and complex analyses of a detailed dataset, but are aware that further studies of thinning and fire severity relationships are warranted, because some variables could not be included in our modeling. For example, there would be merit in quantifying the effects of thinning across broader landscapes and its influence on patterns of fire spread. In addition, our study focused on a generally systematic approach to stand thinning (where a broadly similar amount of biomass would likely have been removed). Further studies might examine thinning and fire severity relationships under different thinning regimes (e.g., a range of different amounts and spatial patterns of biomass removal). It may also be useful to examine the effects of the amount of time elapsed since thinning, although a different kind of analysis would be required to that employed in this study where the aim was to compare fire severity of thinned stands with those that had never been thinned.

4.3 Management implications

The findings of our study suggest that the case for widespread thinning as a strategy for reducing fire severity (see AFPA, 2020) appears to have limited empirical support. Indeed, in some cases, it may exacerbate fire risks. An exception was the impact of thinning in young (20- to 40-year-old) stands of mixed species forest subject to a Crown Burn, although thinning in more mature stands (∼ 70 years old) increased the probability of Crown Burn (Figures 5a and S3).

ACKNOWLEDGMENTS

We acknowledge the Traditional Custodians of the Land upon which our study was based: the Gunaikurnai, Taungurong, and Wurundjeri Peoples. Data sources are detailed in the Supplementary Information Summary S1. All data are publicly available and cited in the references.

    AUTHOR CONTRIBUTIONS

    CT was involved with the conceptualization; methodology; data curation; investigation; formal analysis; writing––original draft; writing––review and editing. WB was involved with the methodology; investigation; formal analysis; and writing––original draft. DBL was involved with the conceptualization; writing––original draft; writing––review and editing; and supervision.

    CONFLICT OF INTEREST

    The authors have no conflict of interest.

    DATA AVAILABILITY STATEMENT

    All data are from publicly available sources and are referenced in the paper and the Supplementary Information.